| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 School of Kinesiology, University of Illinois, Chicago, Illinois 60608; 4 Departments of Bioengineering and Physical Therapy, University of Illinois, Chicago, Illinois 60608; 2 Neuromuscular Research Center, Boston University, Boston, Massachusetts 02215; 3 Department of Neurological Sciences, Rush-Presbyterian St. Luke's Medical Center, Chicago, Illinois 60612
Submitted 7 October 2003; accepted in final form 11 January 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
; Berthoz 2000), and it has been at the forefront of conceptual development within the framework of internal models over the last decade (reviewed in e.g., Kawato 1999; Miall and Wolpert 1996; Wolpert and Flanagan 2001).
The CNS also has at its disposal sensory feedback of different modalities that can be used to correct for discrepancies between the desired and actual movement outcome. The delays in the visual and proprioceptive pathways, however, present a major problem for feedback control of movement. Even the fastest-acting segmental reflexes have a loop delay of 
30–50 ms which might be enough to destabilize the limb (Prochazka and Trend 1988). Despite the delays, proprioceptive feedback is crucial for the control of movement. For example, it has been suggested that the increased sensitivity of the proprioceptive feedback response to early trajectory errors underlies the movement impairment in Huntington's disease (Smith et al. 2000). Experiments in deafferented subjects have demonstrated loss of interjoint coordination in multi-joint movements (Sainburg et al. 1993, 1995) and decreased accuracy in single-joint movements (Nougier et al. 1996; Sanes et al. 1984). However, it remains unclear exactly how proprioceptive feedback operates during fast movement.
Previous experimental studies in which torque perturbations or unexpected loads have been applied at various times before or during movement have not provided a definitive answer. Torque pulses applied <50 ms before or after the agonist electromyographic (EMG) onset did not elicit a short-latency EMG response (Brown and Cooke 1981; Hallett et al. 1975; Hayashi et al. 1990). When perturbations were applied at or after the beginning of the movement, the EMG responses were observed at latencies ranging from 25 to >100 ms from the time that the limb started moving or the load was changed. Based on the latency of the EMG responses, it was concluded that the segmental reflexes were active during movement (Bennett 1993; Latash 1994; Lee et al. 1986; Sanes 1986; Smeets et al. 1990), were not active during movement (Day and Marsden 1982; Wadman et al. 1979), or were initially suppressed and facilitated later during movement (Gottlieb 1996). The EMG responses were attributed to long-latency reflexes (Day and Marsden 1982) or to short-latency segmental reflexes. In the latter case, it was argued that the EMG responses appeared late because the segmental reflexes were being gated by the supraspinal input for a fixed period of time after the agonist EMG onset (Brown and Cooke 1981; Hallett and Marsden 1979). It has been also suggested that the EMG responses are triggered when the velocity error, defined as the difference between the expected and unexpected velocities, exceeds a fixed threshold (Smeets et al. 1990). However, we have recently shown that the EMG responses to an unexpected load were not related to a fixed velocity-error threshold (Shapiro et al. 2002).
The published results do not lend themselves to a straightforward comparison. The experimental protocols included torque pulse perturbations (Cooke 1980), unexpected changes in bias torque (Lee et al. 1986; Marsden et al. 1976b), movement arrest or release (Hallett and Marsden 1979; Ives et al. 1993, 1999; Wadman et al. 1979), constant position error (Bennett 1993), or changes in the load inertia or viscosity (Day and Marsden 1982; Gottlieb 1996; Latash 1994; Sanes 1986; Smeets et al. 1990). In some of the studies, the onset of the EMG responses was given with respect to the agonist EMG onset. In the other studies, the latency of EMG responses was measured from the first detectable change in movement kinematics. However, an unexpected change in the load parameters affected the movement kinematics only after the limb started moving. To compare the published results, it is necessary to account for at least a 20-ms delay between the agonist EMG onset and detectable movement onset (Corcos et al. 1992). In addition, in some of the experiments the movement onset was further delayed because the limb started moving only after the joint torque exceeded a threshold (Hallett et al. 1975; Lee et al. 1986; Marsden et al. 1976a; Smeets et al. 1990). Once the delay between the agonist EMG onset and change in movement kinematics onset is taken into account, it can be concluded that in all of the experiments described in the preceding text, the EMG responses were never observed earlier than 100–200 ms after the agonist EMG onset. The question therefore remains what determines the timing of the EMG responses?
We suggest that it is the time course of the central regulation of the segmental reflex gains that determines the timing of EMG responses to the external mechanical stimuli. It has been shown that the magnitude of the short-latency EMG responses to perturbations increased just before the agonist EMG onset, sharply decreased right after the EMG onset for 100–150 ms, then increased again and returned to a baseline level at the end of movement (Gottlieb and Agarwal 1980; Soechting et al. 1981). It was suggested that the magnitude of the EMG responses changed because of the central regulation of the segmental reflex gains. The 25– to 50-ms short-latency EMG responses to the torque perturbations delivered >50 ms prior to the agonist EMG onset reported in Adamovich et al. (1997), Brown and Cooke (1986), Cordo (1990), and Koshland and Hasan (2000) may be explained by the initial increase of the segmental reflex gains. On the other hand, the subsequent decrease in the gains may explain the absence of the short-latency EMG responses at the beginning of the movement.
We hypothesize that the duration of the central inhibition of the segmental reflex feedback in the beginning of movement depends on the expected movement time. In our previous paper (Shapiro et al. 2002), we reported the results of an experiment in which a viscous load was unexpectedly applied in two movement tasks with the same expected movement time. The tasks included fast elbow flexion movements made over a 30° distance against an expected heavy inertial load (task 30H) and over a 50° distance against an expected light inertial load (task 50L). We investigated whether the onset of the EMG responses was dependent on the timing or magnitude of the velocity error. We observed no difference in the onset of the EMG responses in the two tasks. We concluded that the onset of the EMG responses was not determined by the onset or a fixed threshold of the velocity error and was the same in the two tasks because these tasks had the same expected movement time. The observed result, however, could also indicate that the segmental reflexes were inhibited for a fixed period of time after the movement onset. To distinguish between these two possibilities, in this study, we analyzed four movement tasks. In addition to previously analyzed tasks 30H and 50L, we analyzed movements made over a 30° distance against the light load (task 30L) and 50° distance against the heavy inertial load (task 50H). Thus the experimental protocol included the short and long movements made against the light load (tasks 30L and 50L) and heavy load (tasks 30H and 50H). The movement duration increased with an increase in the distance and/or load. We predicted that when viewed across all four tasks the onset of the EMG responses to the unexpected load would be delayed with an increase in the expected movement duration. A preliminary report has been published in abstract form (Shapiro et al. 2003).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Eight neurologically normal volunteers (5 males and 3 females, aged 20–44 yr) participated in the study. All subjects gave informed consent according to a University approved protocol.
Apparatus
The apparatus included a motor-driven manipulandum, which consisted of a metal bar freely rotating in a horizontal plane around a pivot centered at the elbow joint. The seated subject abducted the shoulder 90° and rested the forearm on the bar with the elbow joint aligned with the pivot. A computer monitor in front of the subject showed markers indicating the initial and target positions and a cursor indicating the current joint angle. The pivot was attached to a shaft of a torque motor (JR25 ServoDisk, Kollmorgen). The torque motor simulated an additional inertial or viscous load by generating resistive torque proportional to the movement acceleration or velocity. The motor torque was measured by a torque transducer inserted between the bar axis and motor shaft, joint angle and acceleration signals were transduced, the velocity signal was obtained from the angle signal using an analog differentiating circuit. All mechanical signals were low-pass filtered using an analog third-order Bessel filter with 60 Hz cutoff frequency. Surface EMGs in biceps, brachioradialis, long and lateral heads of triceps were recorded using Delsys electrodes (gain: 1,000, 20–450 Hz built-in analog band-pass filter). All data were digitized at 1,000 samples/s (National Instruments). The mechanical channels were digitally low-pass filtered with a two-way second-order Butterworth filter with 20 Hz cutoff frequency, the EMG were digitally full-wave rectified and then low-pass filtered with the two-way second-order Butterworth filter with 50 Hz cutoff frequency.
Experimental protocol
Subjects were asked to align the cursor with an initial position marker, wait for a GO beep, and make a fast elbow flexion movement to the target, remain at the target until the END beep, and slowly move back to the initial position. The target width was 3°. When required, the computer turned on an additional load simultaneously with the GO beep and turned off the load with the END beep. The instruction was "move as fast as possible, try to reach the target but do not correct if you miss it."
The experimental protocol consisted of four movement tasks: a 30° movement with a light inertial load of 0.1 kgm2 of the manipulandum itself (30L); a 30° movement with a heavy inertial load that was a combination of the manipulandum inertia and an additional inertia of 0.18 kgm2 simulated by the torque motor (30H); a 50° movement with the light load (50L); and a 50° movement with the heavy load (50H). We expected that an increase in movement distance and/or load would increase the movement time and time to peak velocity (TPkV) and prolong the first agonist burst (Gottlieb et al. 1989; Mustard and Lee 1987; Pfann et al. 1998).
In each task, subjects first made 30–50 practice trials. After the first 10 practice trials and during the subsequent test series, the cursor was extinguished when the velocity exceeded +10°/s at the beginning of movement and re-appeared when it fell below +10°/s at the end of movement. Thus the computer screen did not provide the visual feedback during movement but showed the final position at the end of each trial. After the practice series, the subject made a test series of 60 movements. In 12 pseudorandomly chosen trials, the motor applied an unexpected viscous load of 3.6 Nms/rad. At least two consecutive trials against the expected load followed a trial against the unexpected load to prevent adaptation (Weeks et al. 1996). In the light-load tasks, 30L and 50L, the viscous load was unexpectedly added. In the heavy-load tasks, 30H and 50H, the viscous load was added concurrent with the removal of the additional inertial load. This was done by switching the feedback from the acceleration signal to the velocity signal. Thus a combination of the viscous load and manipulandum inertia was the same in all four tasks. The unexpected load altered the movement trajectory and produced a velocity error. This experimental design produced unexpectedly loaded movements with different velocity errors in the four tasks.
Analysis of muscle EMG
In the analysis of 48 trials against the expected load in each task, we excluded atypical trials. In most cases, these were 2 or 3 of the shortest, longest, slowest, and fastest trials, usually 8–12 trials total. All 12 trials against the unexpected viscous load were accepted. The data were aligned on the biceps (agonist) EMG onset. The point of alignment was set as t = 0.
Because of the high variability of the EMG signal, the task of identifying the EMG responses to the unexpected load presents a major methodological problem. We dealt with this problem as follows. We determined the time intervals when the difference between the EMG signals in the movements against the expected and unexpected loads was statistically significant (Shapiro et al. 2002). For each muscle, two sets of the EMG values from the trials against the expected (n = 36–40) and unexpected loads (n = 12) were compared using Satterthwaite's modified t-test (Armitage and Berry 1994). The test was repeated for every other sample point, every 2 ms, for a time interval t = 0–0.5 s to generate a P value time series. The EMGs in movements against the expected and unexpected loads were considered statistically different if P fell <0.05. This procedure is illustrated in Fig. 1.
|
10 ms. Thus the EMG difference within interval I in Fig. 1 was not considered the EMG response because its duration was <10 ms. The second requirement was that the sign of an EMG difference, i.e., an increase or decrease in the EMG, had to be consistent with the sign of the preceding velocity error. In tasks 30L and 50L, the resistive torque was larger than expected from the very beginning of the movement, so the movement was slower and the velocity error was always positive (described in detail in RESULTS). The corresponding EMG responses had to be an increase in the agonist EMGs and/or decrease in the antagonist EMGs. In tasks 30H and 50H, however, the velocity error was initially negative and then changed sign (described in detail in RESULTS). The EMG responses to the initial negative velocity error had to be a decrease in the agonist EMG and/or increase in the antagonist EMG. The EMG difference within interval II (Fig. 1) must be a response to the initial negative velocity error. After the velocity error changes sign and becomes positive, the sign of the EMG responses should be the same as in tasks 30L and 50L. Then, the EMG difference within interval III (Fig. 1) must be a response to the positive velocity error.
The third requirement was that the EMG differences of the particular sign should be observed consistently across the subjects, tasks, and muscles. The initial negative velocity error in tasks 30H and 50H elicited a corresponding decrease in the agonist EMG and/or increase in the antagonist EMG only in some of the tasks, some of the subjects, and some of the muscles. These initial EMG differences were analyzed separately. In contrast, the EMG differences elicited by the positive velocity error were found in 124 of 128 instances 8 subjects x 4 muscles x 4 tasks and were consistent across all four tasks, all subjects, and all muscles. These EMG differences were not detected in tasks 30L and 50L in the brachioradialis (subject 6), in task 30L in the long head of triceps (subject 5), and in task 50L in the lateral head of triceps (subject 1).
Analysis of the velocity error
For each subject and each movement task, the velocity error was calculated as the difference of the averaged velocity time profiles in the movements against the expected and unexpected loads. In many instances, the velocity error initially exhibited small deviations around zero with the magnitude <2°/s before the onset of a large and rapid change. After the onset, the velocity error signal changed by >2°/s between the consecutive samples separated by 2 ms. In tasks 30L and 50L, the velocity error was positive from the beginning of the movement and its onset was taken when it first exceeded +2°/s. In tasks 30H and 50H, the onset of the initial negative velocity error was taken when it first fell below –2°/s. Afterward, the velocity error changed sign and exceeded +2°/s within 2 ms of crossing zero. For consistency, the onset of the positive velocity error in all four tasks was taken when it first exceeded +2°/s.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
EMG response onset in the four movement tasks
The main experimental result of this study is that the onset of the EMG responses to the unexpected viscous load is delayed with the expected movement time. In our experiment, the distance and load manipulations were used as the indirect means to increase the expected movement duration or TPkV (Pfann et al. 1998). We used TPkV as an indicator of the movement duration because it was the least ambiguous parameter to determine. Both the load and distance had significant effects on the TPkV (Table 1). The data in Fig. 2 are taken from one representative subject and show that the EMG response onset was progressively delayed as the expected movement time increased with an increase in the movement distance and/or expected load (Fig. 2, the EMG response onset is indicated, 
). In all four tasks, the EMG responses were an increase in the magnitude of the agonist EMG (biceps and brachioradialis) and a decrease in the magnitude of the antagonist EMG (long and lateral heads of triceps; Fig. 2, Biceps, triceps panels).
|
|
|
Muscle EMG responses elicited by positive and negative velocity errors
The movement kinematics was affected differently in the tasks with the light (30L and 50L) or heavy (30H and 50H) expected load. This was because in tasks 30L and 50L, the viscous load was unexpectedly added to the manipulandum inertia, whereas in tasks 30H and 50H, the viscous load was substituted for the expected additional inertial load generated by the torque motor. In tasks 30L and 50L, the resistive torque was larger than expected for the entire movement duration (Fig. 2, tasks 30L and 50L, torque panels). The resulting movement was slower than expected so that the velocity error was positive from the beginning of the movement. The peak velocity (PkV) was lower than expected and was reached earlier (tasks 30L and 50L in Figs. 2, velocity panels, and 4, PkV, TPkV). The peaks of acceleration and deceleration were lower than expected [tasks 30L and 50L in Figs. 2, acceleration panels, and 4, peak acceleration (PkAcc), peak deceleration (PkDec)]. Because the unexpectedly loaded movement was slower than expected, the agonist muscles were shortened less than expected, whereas the antagonist muscles were stretched less than expected. The observed increase in the agonist EMG and decrease in the antagonist EMG should have been caused by the relatively less shortening of the agonist and stretching of the antagonist muscles (tasks 30L and 50L in Fig. 2, biceps, triceps panels).
In tasks 30H and 50H, the unexpected resistive torque was proportional to the velocity and initially lagged behind the expected torque that would have been proportional to the acceleration (tasks 30H and 50H in Fig. 2, torque panels). Consequently, the limb was initially moving faster than expected reaching higher PkAcc (tasks 30H and 50H in Figs. 2, acceleration and velocity panels, and 4, PkAcc). The PkV in the unexpectedly loaded movements was not significantly different from the expected PkV (tasks 30H and 50H in Fig. 4, PkV) but was reached significantly earlier than expected (tasks 30H and 50H in Fig. 4, TPkV). After the initial period of relative acceleration, the resistive torque became larger than expected and caused a vigorous deceleration so that the movement became slower than expected (tasks 30H and 50H in Figs. 2, torque, acceleration, and velocity panels, and 4, PkDec). Consequently, the velocity error was negative during the initial acceleration of the limb and then became positive (tasks 30H and 50H in Fig. 2, velocity panels). Similarly to tasks 30L and 50L however, the earliest consistent EMG responses in tasks 30H and 50H were an increase in the magnitude of the agonist EMG and decrease in the magnitude of the antagonist EMG in all muscles and all subjects. These changes in the magnitude of the muscle EMGs must have been elicited by the positive velocity error or relative slowing of the movement.
|
|
EMG response onset was not determined by the velocity error onset or threshold
The increase in the expected load in our experiment increased the TPkV (Table 1) and also delayed the onset of the positive velocity error [load effect F(1,7) = 404.56, P < 0.0001; see also Fig. 5A, 
and 
]. This delay was produced because the unexpected viscous load was substituted for the heavy inertial load in tasks 30H and 50H and could explain the observed delay in the EMG response onset. In this case, the EMG responses should have been elicited at a fixed latency with respect to the velocity error onset. We calculated the latency of the EMG response with respect to the velocity error onset for the biceps muscle because it showed the EMG responses in all subjects and all tasks. This latency significantly decreased in the heavy load tasks [load effect F(1,7) = 29.97; P < 0.001; see also Fig. 5A, 
and 
]. Therefore we conclude that the observed delay in the EMG response onset could not be explained by the delay in the velocity error onset.
|
) that in all movement tasks, the EMG responses appeared at a short latency after the velocity error exceeded a fixed threshold. In this case, the magnitude of the velocity error 30 ms prior to the onset of the biceps EMG response (VE30) should not have changed with the expected movement task. The VE30, however, significantly increased with the increase in distance [distance effect F(1,7) = 8.65, P < 0.02; Fig. 5B], and the effect of load came close to reaching significance [F(1,7) = 4.71, P < 0.067; Fig. 5B]. We conclude that the observed delay in the EMG response onset could not be explained by the effects of the load and distance manipulations on the velocity error. Next, we test whether the EMG response timing could be explained by the expected movement time.
Effect of prediction on the muscle EMG
We analyzed the correlation between the timing of biceps EMG response and TPkV, which provides a purely temporal characteristic of the task. To pool together the data for all eight subjects, we calculated the change in the onset of biceps EMG response (
tEMG) with an increase in the expected load and/or distance. The tEMG for tasks 30H, 50L, and 50H was calculated by subtracting the EMG response onset in 30L from the corresponding values in the remaining three tasks. Similarly, the change in TPkV (TPkV) for tasks 30H, 50L, and 50H was calculated by subtracting the TPk in 30L from the corresponding values of TPk in the remaining three tasks. As can be seen in Fig. 6, the tEMG and TPkV are significantly correlated (2-tailed P < 0.001).
|
|
150–200 ms. After that the muscles apparently relaxed although the subject undershot the target in these tasks (Fig. 7, angle panels). When viewed across the subjects, the endpoint error, or the difference between the final positions in movements against the expected and unexpected loads, was highly variable. Seven subjects undershot the target in at least three of four tasks, while one subject either reached or overshot the target (Table 3, subject 4). The endpoint error was not significantly affected by the expected load [F(1,7) = 1.46; P < 0.27] or distance [F(1,7) = 0.21; P < 0.67]. Therefore we analyzed the endpoint error by pooling together the data for all subjects and all tasks (n = 32). The endpoint error was a statistically significant 5.1° undershoot (P < 0.0001, 1-group 2-tailed t-test).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
While a number of possible mechanisms could generate an EMG response at a latency of >100 ms from the velocity error onset, a short latency of 30 ms would unambiguously point to segmental reflexes. The shortest latency was 34 ms observed in two subjects in task 30H, and in a number of instances, the EMG responses in tasks 30H and 50H were detected at a latency of 40 ms (Table 4). We conclude that in these particular cases the EMG responses were mediated by segmental reflexes. At this point, we could hypothesize that in all other cases the EMG responses were generated by some mechanism(s), other than the segmental reflexes, operating at longer latencies with respect to the velocity error onset. However, all four tasks in our experiment were qualitatively similar, i.e., moving an inertial load from an initial position to the target as fast as possible, and we argue that the control processes during these movements were similar as well. Therefore we suggest that the EMG responses in all four tasks were due to segmental reflexes, but their timing was not determined solely by the nerve conduction delays along the segmental reflex pathways. This hypothesis brings us to a seeming paradox that the "short latency" segmental response, which was driven by the velocity error, occurred in many instances >100 ms from the velocity error onset (Table 4). This timing could not be explained by relating the EMG response to a fixed velocity error threshold 30 ms prior to the EMG response onset (Fig. 5B). Although the time 30 ms was arbitrarily picked to test whether the EMG response was elicited at a short latency from a velocity error threshold, the biceps EMG response in tasks 50L and 30H did not appear to be related to a fixed velocity error threshold at any latency (see also Fig. 2C in Shapiro et al. 2002). In our view, the most parsimonious explanation for our findings is that a signal that triggers the muscle EMG response is the product of the velocity error and the centrally controlled gains of the segmental reflexes.
|
Supraspinal regulation of the segmental reflex gains
The idea that the segmental reflexes are under the descending control is supported by considerable experimental evidence. For example, the stretch reflex during movement can be modified by the central command circuits in most animal groups (reviewed in Clarac et al. 2000). Reflex activity during movement can be modified by central inputs to spinal interneurons (reviewed in Fetz et al. 2000; Jankowska 2001) and presynaptic inhibition (Aymard et al. 2001; Meunier and Pierrot-Deseilligny 1989; reviewed in Rudomin and Schmidt 1999). The current view is that supraspinal inputs organize the segmental circuitry into flexible functional units according to the motor task (recently reviewed in Bosco and Poppele 2001; Hultborn 2001). It has been shown that the CNS actively tunes the magnitude and even sign of the reflex responses as part of a "central set" (Cordo 1990; Evarts 1975; Hore et al. 1990; Lacquaniti and Maioli 1989; Winstein et al. 2000).
It could also be argued that the EMG responses are absent during the initial phase of a fast movement because of decreased muscle spindle sensitivity rather than supraspinal modulation of segmental feedback gains. The sensitivity of muscle spindles is highest for very small stretches and declines if the spindle is stretched further (Hasan 1983; Kakuda 2000). However, the CNS may adjust the muscle spindle sensitivity by tuning the fusimotor system to the expected motor task (Ellaway et al. 2002; Jones et al. 2001; Kakuda et al. 1996; Schieber and Thach 1980). In our experiment, indirect evidence that the proprioceptive information was available to the CNS from the very beginning of the movement was provided by the subjects' reporting. When asked to describe the unexpected load, all of the subjects stated that in tasks 30H and 50H "the manipulandum felt lighter in the beginning, and the resistance appeared only at the end of the movement." The proprioceptive feedback in the beginning of the movement could be provided by cutaneous receptors and/or muscle and joint receptors. Cutaneous sensitivity is decreased during the initial part of movement (Angel and Malenka 1982; Williams and Chapman 2002); this suggests that the information about the change in load was provided by the muscle and joint receptors from the very beginning of the movement. The output of the muscle spindle afferents is routed to the segmental pathways as well as the ascending pathways. The synaptic effectiveness of the segmental and ascending collaterals of individual muscle spindle afferents can be independently affected by central mechanisms (Lomeli et al. 1998; Rudomin 2002). Thus the same proprioceptive information can be used differently for the segmental feedback control and tasks performed by higher brain structures, such as state estimation (Cordo et al. 1994; Wolpert et al. 1995), learning a novel load (Thoroughman and Shadmehr 1999), or updating an internal model of the limb (Ghez and Sainburg 1995; Vercher et al. 2003).
Interaction of the central controller and segmental feedback during fast movement
The mechanism of independent control of the ascending and spinal afferent inputs coupled with the ability to dynamically organize the spinal reflex circuitry provides the CNS with significant flexibility in countering destabilizing effects of the feedback delays. The central and segmental feedback controllers can be viewed as master and slave. The master is the local controller in the brain and slave is the remote controller in the spinal cord. The local controller may include an inverse dynamic model that generates a motor command and a forward model that predicts the sensory consequences of this motor command and provides fast internal feedback (Jordan and Rumelhart 1992; Wolpert and Kawato 1998). The predicted sensory signal within the local controller in the brain can be delayed to improve the system stability by canceling the reafference (Miall et al. 1993). This system may become unstable if the internal model is inaccurate and the predicted and actual afferent signals do not match. The destabilizing effect of long transmission delays between the local and remote controllers can be dealt with by using the combinations of the force and velocity signals to communicate between the local and remote controllers (Massaquoi and Slotine 1996; Niemeyer and Slotine 1991). The response to perturbations in this case may be sluggish if the transmission delays are long, so this approach might be best used for the control of slower movements.
One possibility is that all feedback loops, including the spinal feedback within the remote controller, operate continuously during movement. This control structure has been explored in the modeling studies of reaching movements (Bhushan and Shadmehr 1999; Wang et al. 2001). Our results indicate, however, that the feedback control during fast point-to-point movements is activated for a short time in the second half of the movement. This is consistent with the idea that strong segmental reflex feedback operates during a limited interval during movement so that the instability is not expressed (Rack 1981). In our experiment, the muscles relaxed after a brief burst of the feedback-generated activity so that the unexpectedly loaded movements were shorter in most cases. Thus the segmental reflex feedback was not acting as a simple servo system that would eventually bring the limb to the target. When the movement was slowed by the unexpected load, the motor system made a short powerful "push" in the target direction using segmental reflex feedback and then relaxed. Mechanisms other than the segmental reflex feedback are likely to be employed if the subject was instructed to compensate for the perturbation (Crago et al. 1976).
To summarize, the control of fast point-to-point movement consists of three phases with the centrally preset timing that depends on the movement task, predictive feedforward control, combination of feedforward control and briefly facilitated feedback, and relaxation. There is also experimental evidence that some features of these components of the control of fast movements are different in males and females (Ives et al. 1993). Our results provide evidence against models of the control of fast movement that include continuous segmental reflex feedback, as in, e.g., (Feldman and Levin 1995; Gribble et al. 1998; Latash 1993; Latash and Gottlieb 1992). It appears that when movement kinematics during fast movement is strongly disrupted by a novel load, the motor system prefers to fail and miss the target rather than risk instability caused by the feedback delays. If this novel load is expected in successive movements, the CNS compensates for it by partially adjusting the feedforward part of the control sequence (Thoroughman and Shadmehr 1999).
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
GRANTS
This work was supported by National Institutes of Health Grants F32 HD-08596, RO1-AR-33189, RO1-AR-44388, R01-NS-28127, and RO1-NS-40902.
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: M. B Shapiro, Dept. of Movement Sciences (MC 994), University of Illinois at Chicago, 808 South Wood St., 690 CME, Chicago, IL 60612 (E-mail: mshapi2{at}uic.edu).
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Angel RW and Malenka RC. Velocity-dependent suppression of cutaneous sensitivity during movement. Exp Neurol 77: 266–274, 1982.[CrossRef][ISI][Medline]
Armitage P and Berry G. Statistical Methods in Medical Research: Oxford, UK: Blackwell Science, 1994.
Aymard C, Baret M, Katz R, Lafitte C, Penicaud A, and Raoul S. Modulation of presynaptic inhibition of la afferents during voluntary wrist flexion and extension in man. Exp Brain Res 137: 127–131, 2001.[CrossRef][ISI][Medline]
Bennett DJ. Electromyographic responses to constant position errors imposed during voluntary elbow joint movement in human. Exp Brain Res 95: 499–508, 1993.[ISI][Medline]
Bernstein NA. The Coordination and Regulation of Movement. Oxford, UK: Pergamon, 1967.
Berthoz A. The Brain's Sense of Movement. Cambridge, MA: Harvard Univ. Press, 2000.
Bhushan N and Shadmehr R. Computational nature of human adaptive control during learning of reaching movements in force fields. Biol Cybern 81: 39–60, 1999.[CrossRef][ISI][Medline]
Bosco G and Poppele RE. Proprioception from a spinocerebellar perspective. Physiol Rev 81: 539–568, 2001.
Brown SH and Cooke JD. Responses to force perturbations preceding voluntary human arm movements. Brain Res 220: 350–355, 1981.[CrossRef][ISI][Medline]
Brown SH and Cooke JD. Initial agonist burst is modified by perturbations preceding movement. Brain Res 377: 311–322, 1986.[CrossRef][ISI][Medline]
Clarac F, Cattaert D, and Le Ray D. Central control components of a "simple" stretch reflex. Trends Neurosci 23: 199–208, 2000.[CrossRef][ISI][Medline]
Cooke JD. The organization of simple, skilled movements. In: Tutorials in Motor Behaviour, edited by Stelmach G and Requin J. Amsterdam: Elsevier, 1980.
Corcos DM, Gottlieb GL, Latash ML, Almeida GL, and Agarwal GC. Electromechanical delay: an experimental artifact. J Electromyogr Kinesiol 2: 59–68, 1992.
Cordo PJ. Kinesthetic control of a multijoint movement sequence. J Neurophysiol 63: 161–172, 1990.
Cordo P, Carlton L, Bevan L, Carlton M, and Kerr GK. Proprioceptive coordination of movement sequences: role of velocity and position information. J Neurophysiol 71: 1848–1861, 1994.
Crago PE, Houk JC, and Hasan Z. Regulatory actions of human stretch reflex. J Neurophysiol 39: 925–935, 1976.
Day BL and Marsden CD. Accurate repositioning of the human thumb against unpredictable dynamic loads is dependent upon peripheral feedback. J Physiol 327: 393–407, 1982.
Ellaway P, Taylor A, Durbaba R, and Rawlinson S. Role of the fusimotor system in locomotion. Adv Exp Med Biol 508: 335–342, 2002.[ISI][Medline]
Evarts EV. The Third Stevenson Lecture. Changing concepts of central control of movement. Can J Physiol Pharmacol 53: 191–201, 1975.[ISI][Medline]
Feldman AG and Levin MF. The origin and use of positional frames of reference in motor control. Behav Brain Sci 18: 723–806, 1995.[ISI]
Fetz EE, Perlmutter SI, and Prut Y. Functions of mammalian spinal interneurons during movement. Curr Opin Neurobiol 10: 699–707, 2000.[CrossRef][ISI][Medline]
Ghez C and Sainburg R. Proprioceptive control of interjoint coordination. Can J Physiol Pharmacol 73: 273–284, 1995.[ISI][Medline]
Gottlieb GL. On the voluntary movement of compliant (inertial-viscoelastic) loads by parcellated control mechanisms. J Neurophysiol 76: 3207–3229, 1996.
Gottlieb GL and Agarwal GC. Response to sudden torques about ankle in man. III. Suppression of stretch-evoked responses during phasic contraction. J Neurophysiol 44: 233–246, 1980.
Gottlieb GL, Corcos DM, and Agarwal GC. Organizing principles for single-joint movements. I. A speed-insensitive strategy. J Neurophysiol 62: 342–357, 1989.
Gribble PL, Ostry DJ, Sanguineti V, and Laboissiere R. Are complex control signals required for human arm movement? J Neurophysiol 79: 1409–1424, 1998.
Hallett M and Marsden CD. Ballistic flexion movements of the human thumb. J Physiol 294: 33–50, 1979.
Hallett M, Shahani BT, and Young RR. EMG analysis of stereotyped voluntary movements in man. J Neurol Neurosurg Psychiatry 38: 1154–1162, 1975.
Hasan Z. A model of spindle afferent response to muscle stretch. J Neurophysiol 49: 989–1006, 1983.
Hayashi R, Becker WJ, and Lee RG. Effects of unexpected perturbations on trajectories and EMG patterns of rapid wrist flexion movements in humans. Neurosci Res 8: 100–113, 1990.[CrossRef][ISI][Medline]
Hore J, McCloskey DI, and Taylor JL. Task-dependent changes in gain of the reflex response to imperceptible perturbations of joint position in man. J Physiol 429: 309–321, 1990.
Hultborn H. State-dependent modulation of sensory feedback. J Physiol 533: 5–13, 2001.
Ives JC, Abraham L, and Kroll W. Neuromuscular control mechanisms and strategy in arm movements of attempted supranormal speed. Res Q Exerc Sport 70: 335–348, 1999.[ISI][Medline]
Ives JC, Kroll WP, and Bultman LL. Rapid movement kinematic and electromyographic control characteristics in males and females. Res Q Exerc Sport 64: 274–283, 1993.[ISI][Medline]
Jankowska E. Spinal interneuronal systems: identification, multifunctional character and reconfigurations in mammals. J Physiol 533: 31–40, 2001.
Jones KE, Wessberg J, and Vallbo AB. Directional tuning of human forearm muscle afferents during voluntary wrist movements. J Physiol 536: 635–647, 2001.
Jordan MI and Rumelhart DE. Forward models: Supervised learning with a distal teacher. Cognit Sci 16: 307–354, 1992.
Kakuda N. Response of human muscle spindle afferents to sinusoidal stretching with a wide range of amplitudes. J Physiol 527: 397–404, 2000.
Kakuda N, Vallbo AB, and Wessberg J. Fusimotor and skeletomotor activities are increased with precision finger movement in man. J Physiol 492: 921–929, 1996.[ISI][Medline]
Kawato M. Internal models for motor control and trajectory planning. Curr Opin Neurobiol 9: 718–727, 1999.[CrossRef][ISI][Medline]
Koshland GF and Hasan Z. Electromyographic responses to a mechanical perturbation applied during impending arm movements in different directions: one-joint and two-joint conditions. Exp Brain Res 132: 485–499, 2000.[CrossRef][ISI][Medline]
Lacquaniti F and Maioli C. The role of preparation in tuning anticipatory and reflex responses during catching. J Neurosci 9: 134–148, 1989.[Abstract]
Latash ML. Control of fast elbow movement: a study of electromyographic patterns during movements against unexpectedly decreased inertial load. Exp Brain Res 98: 145–152, 1994.[ISI][Medline]
Latash ML. Control of Human Movement. Urbana, IL: Human Kinetics, 1993.
Latash ML and Gottlieb GL. Virtual trajectories of single-joint movements performed under two basic strategies. Neuroscience 47: 357–365, 1992.[CrossRef][ISI][Medline]
Lee RG, Lucier GE, Mustard BE, and White DG. Modification of motor output to compensate for unanticipated load conditions during rapid voluntary movements. Can J Neurol Sci 13: 97–102, 1986.[ISI][Medline]
Lomeli J, Quevedo J, Linares P, and Rudomin P. Local control of information flow in segmental and ascending collaterals of single afferents. Nature 395: 600–604, 1998.[CrossRef][Medline]
Marsden CD, Merton PA, and Morton HB. Servo action in the human thumb. J Physiol 257: 1–44, 1976a.
Marsden CD, Merton PA, and Morton HB. Stretch reflex and servo action in a variety of human muscles. J Physiol 259: 531–560, 1976b.
Massaquoi SG and Slotine JJ. The intermediate cerebellum may function as a wave-variable processor. Neurosci Lett 215: 60–64, 1996.[CrossRef][ISI][Medline]
) and unexpected (
) loads in the 4 tasks. Means ± SD (error bars) for 8 subjects are shown. —, statistically significant differences in the parameter values (*, P < 0.001; **, P < 0.05).
