Comparison of motor copy and targeted biofeedback training techniques for restitution of upper extremity function among patients with neurologic disorders.

Comparison of Motor Copy and Targeted Biofeedback Training Techniques for Restitution of Upper Extremity Function Among Patients with Neurologic Disorders Most physical therapy interventions for the treatment of unilateral limb dysfunction among patients with neurologic disorders are directed exclusively toward the involved limb. This approach is somewhat surprising because efferent pathways originating in the core of the ventromedial brain stem project bilaterally and, as a result, can influence motor output to both upper or lower extremities. [1,2] Evidence from experiments performed in rhesus monkeys indicates that cortical fibers terminate in the same location as the ventromedial group and hence can influence motoneurons of distal extremity muscles. [3,5] Using a split-brain rhesus monkey model, Brinkman and Kuypers have demonstrated that each half of the brain has full control of the movements of the contralateral extremity and also controls proximal and complex movements of the ipsilateral extremity. [6] More specifically, Matsunami and Hamada have shown a population of precentral motoneurons that discharge during bilateral movements involving finger, wrist, or arm activities in monkeys. [7] Collectively, these studies suggest that brain stem and corticospinal neurons are capable of influencing motoneurons supplying upper extremity musculature bilaterally. Research involving bilateral motor systems, therefore, provides an avenue and precedent upon which a new rehabilitation technique could be based.

Surprisingly, only one clinical study has made use of both bilateral descending motor pathways and muscle feedback. [8] This study used a computer interface procedure in which integrated EMG activity generated from the uninvolved facial musculature of a patient with Bell's palsy was matched with integrated muscle output from the contralateral musculature in a patient who two years previously had undergone a spinal accessory-facial nerve anastomosis. As a result of attempting to match muscle output of the involved side of the face with that of the uninvolved side, symmetrical and cosmetically acceptable facial movements were achieved.

Audiovisual representations of muscle activity could conceivably be provided to patients from homologous muscle groups of the upper or lower extremities. By using different gains or sensitivities from each EMG amplifier (ie, high gain for paretic muscles and low gain for normal muscles), patients could self-train to match outputs on a video display. We refer to this unique mode of biofeedback intervention as "motor copy" (MC) because clients are attempting to copy representations of output from a normal (uninvolved) muscle with output from the homologous, weak muscle. In most conventional dual-channel EMG biofeedback training, patients do not monitor homologous muscles but rather try to reduce activity in a spastic muscle and recruit activity in the antagonist muscle. Both these muscles are considered to be "targeted" for differential training, hence the generic term targeted training (TT). [9] The clinical significance of the MC procedure resides partly in its potential cost-effectiveness because once a patient is set up for muscle monitoring, much of the bilateral training could be undertaken without the need of a therapist presiding. Patients also quickly recognize that they can use the uninvolved muscle to "activate" a paretic muscle over which they previously had no control.

The present study determined whether this novel technique (MC) was more beneficial in producing appropriate functional and neuromuscular changes in the upper extremities of chronic stroke and head-injured patients than a conventional muscle biofeedback training technique (TT). Based on research results from subhuman primate models, [3,5,7] we hypothesized that 1) the MC procedure would produce EMG and functional changes more appropriate to the reacquisition of shoulder and elbow movement control than the TT procedure and 2) both MC and TT procedures would produce comparable changes in distal upper extremity control. A preliminary report on this study has been published. [10]

Method

Subjects

Twenty stroke patients and 6 head-injured patients participated in this study. All patients had sustained their injuries from one to seven years prior to participation in this study, and all signed informed consent statements. Specific data on these individuals are shown in Table 1. None of these patients had received any rehabilitation during the previous 12 months, and there was no clinical evidence available to suggest that they were continuing to improve. Patients were randomly assigned to either the MC Group (n = 14) or the TT Group (n = 12). All patients received the entire sequence of 30 treatments. They were selected using criteria ascertained in a previous feedback study, [11] which identical factors predicting successful functional return among chronic stroke patients. Inclusion criteria included the ability to comprehend and respond appropriately to instructions while being trained with EMG biofeedback equipment; availability to attend all evaluation, treatment, and follow-up evaluation sessions; the ability to at least initiate voluntary wrist and finger extension and thumb abduction movements; and nonparticipation in other rehabilitation programs for the upper extremity. By the 12-month follow-up evaluation, each group had lost three members because of a second stroke, unrelated surgery, or relocation. Thus, the data analysis performed after the 12-month follow-up evaluation was based on 11 MC Group and 9 TT Group patients.

Evaluation and Treatment

Each evaluation, which involved a total of 147 independent variables, consisted of measurement of active ROM, functional tasks based on force or time measures, and integrated EMG levels. A complete listing of functional tasks appears in Table 2. All tasks, except weight to box, extend elbow with weight, and grip strength, were measured in time using stop-watches. The two clinicians (DEL and LAB) measuring these times were tested for interrater reliability by assessing a healthy subject as she performed temporal-specific tasks three times. There were no significant differences between the two clinicians in any timed activity (two-way analysis of variance [ANOVA], p [is greater than] .05). The validity of these tasks was strengthened through the knowledge that no significant group differences existed across tasks or between and within raters when 14 healthy subjects were tested five times over five weeks at the same time of day. The other three tasks were measured with sandbag weights (weight to box and extend elbow with weight) or with a dynamometer (grip strength). Clinician raters showed no significant differences in their measurements. They also produced within-rater and between-raters active ROM measurements that were within [+ or -] 5 degrees for each ROM measurement performed at specified upper extremity joints, movements, and body positions using standard 360-degree goniometers.

Electromyographic activity was assessed through analog-to-digital conversion of the integrated EMG (time constant = 100 msec) sampled every 50 msec using a data-acquisition program and an IBM XT personal computer (*1) (640k of random access memory).

Reliability to monitor a known voltage input was ascertained by passing known voltage signals into IBM (*1) and Hyperion (*2) microprocessors and measuring their amplitudes and rectified integrated values periodically throughout this study. At no time did calculated values from processing of the same known input fluctuate more than 5%. Because of the inherent variability in reproducing statistically reliable quantitative EMG values from spastic muscles, we did not attempt to ascertain reliability of EMG measurements within or between examiners (clinicians).

Evaluations consisted of measuring all 19 functional tasks (Tab. 2), 10 active ROMs, and 18 values of integrated EMG (Tab. 4). Following five baseline evaluations, evaluations 6 to 8 were performed after the 10 treatments to the shoulder, the elbow, and the wrist and hand, respectively. Follow-up evaluations 9 to 11 were undertaken at 3, 6, and 12 months after treatment.

All treatments were performed using the Hyperion 4080 Bioconditioner (*1) dual-channel biofeedback unit. This device has an input impedance of [10.sup.12] [Omega] and a noise level of [is less than]2 [mu]V root mean square. It processes EMG activity across a bandwidth of 20 Hz to 1 kHz and has a frequency response of -3 dB at 20 Hz and 1 kHz, flat within [+ or -] 1 dB over the range.

Motor copy training consisted of monitoring homologous upper extremity muscles in the same proximal-to-distal time sequence as was done for TT; that is, muscle groups were monitored simultaneously for each 10-treatment series, starting in the shoulder region and progressing to the elbow and then to the wrist, fingers, and thumb. The MC training technique is outlined in Table 3. Muscle training under the MC condition necessitated matching by superimposition of homologous muscle outputs. First, the muscle outputs were super-imposed at rest and then when each muscle group was either passively or actively lengthened, usually at a speed approximating a joint angle movement of 20[degrees]/sec as determined from coordinating joint movement to a known oscilloscopic sweep speed. Second, the antagonist muscles were actively shortened simultaneously while patients maintained reduced EMG levels in hyperactive, synergist muscles. Last, the synergists were actively shortened. Matching outputs from each homologous muscle pair was achieved through independently varying the gain on each EMG amplifier. This procedure permitted visualization of linear outputs representing integrated muscle activity levels that could easily be matched. The degree of difficulty was heightened through making amplifier gains more similar, thus necessitating even greater control of output from involved (paretic) muscles to achieve superimposition of line traces on the screen. Traces from either output could be stored on the screen for subsequent matching or erased with each sweep. Both approaches were used depending on patient preference.

Targeted biofeedback training was undertaken following procedures described previously. [9,11] Briefly, using electrode placements oriented across the muscle end-plate region and anatomically referenced for replication of electrode placements, EMG activity from spasti muscles (pectoralis major and upper trapezius muscles about the shoulder; biceps brachii muscle at the elbow; and flexor muscles of the wrist, fingers, and thumb) was downtrained (inhibited) while EMG activity from typically weaker, antagonist muscles (anterior and middle deltoid muscles at the shoulder; triceps brachii muscle at the elbow; and extensor muscles of the wrist, fingers, and thumb) were up-trained (recruited). Thus, patients were trained to reduce hyperactive muscle responses and to recruit paretic muscles. As training progressed to distal muscles, occassional reinforcement of proximal muscles was required through offering verbal or tactile cues or by requiring patients to perform complex tasks that necessitated inclusion of proximal muscle groups. This reinforcement approach did not require repeating EMG biofeedback training to proximal muscle groups.

Data Analysis

To determine whether within-group behavior was similar during acquisition of baseline, treatment, and follow-up data, a Friedman two-way ANOVA was performed on all dependent measurements (EMG, ROM, functional tasks). Mann-Whitney U tests were used for between-group comparisons at the fifth baseline evaluation and following interventions. Statistical significance was established at the .05 level for all changes. Nonparametric statistical analyses were used because of the larger variances created by differences in functional capabilities and EMG acticity so often typical of patients with spasticity.

Results

Baseline Measurements

The only significant baseline changes within groups occurred at the last baseline evaluation and only among a few functional tasks (grip strength, writing name, and lifting a pencil). We, therefore, compared the functional task values at the fifth baseline evaluation to all intervention and follow-up data, ostensibly eliminating any learning effects acquired during earlier baseline evaluations. Only 4 of the 18 EMG measures and 3 of the 19 functional measures showed significant differences between groups at the fifth baseline evaluation. We, therefore, considered the groups to be homogeneous, and a valid comparison between groups following intervention could be made. Furthermore, analysis of within-group baseline measurements segregating side of involvement or head injury versus stroke did not produce significant differences for any dependent measure, further justifying homogeneity among treatment groups.

Within-Group Treatment

Measurements

Motor Copy Group. Table 2 summarizes the 19 functional measures for which statistically significant between-evaluation changes occurred for the MC Group. The tasks are listed according to the joint(s) each task primarily governs. The field headings in Tables 2 and 4 through 8 refer to the possible between-evaluation comparisons from the last baseline evaluation (evaluation 5) through the one-year follow-up evaluation (evaluation 11). For example, "7 v 9" would mean a comparison between the evaluation following treatment of the elbow region (evaluation 7) and the three-month follow-up evaluation (evaluation 11). Table 2 shows that all functional changes were faster movement speeds or greater force measurements, except for the task tracing a circle between evaluations 7 and 8 in which a significant slowing of movement was observed (upward arrow). Among the total 71 significant functional changes between evaluations 5 to 11, 29 (40%) involved the seven shoulder tasks, 26 (37%) were related to the three elbow movements, and 16 (23%) involved the nine wrist and hand tasks.

These changes are depicted graphically in Figure 1, which shows the total number of significant changes possible at evaluations 6 through 11 and those occurring within the MC and TT Groups. The total possible number of significant changes was determined by multiplying the number of functional tasks by the total number of between-evaluation measurements possible at each evaluation. At evaluation 6, for example, only one comparison was made between evaluation 6 and evaluation 5. At evaluation 7, the total number of comparisons possible was 2 (a comparison between evaluation 7 and evaluation 6 and a comparison between evaluation 7 and evaluation 5), and so on. The total number of functional tasks with statistically significant improvements for the MC Group continued to increase across evaluations (Fig. 1, top graph). These progressive increases for the MC Group are reflected and in changes for functional tasks at the shoulder, the elbow (except in evaluation 11), and the wrist and hand (Fig. 1, lower graphs).

Figure 2 shows the distribution of significant changes in ROM across evaluations for the MC Group. Among the total of 15 significant changes, 10 (67%) governed the shoulder and 5 (33%) pertained to the wrist and hand. The total column counts increased over evaluations.

Table 4 shows statistically significant increases (upward arrows) and decreases (downward arrows) in integrated EMG activity across all evaluations for the MC Group. Among the 29 total significant changes, 22 (76%) were seen at follow-up (evaluations 9-11) rather than during treatment. Furthermore, among these changes, only 2 (7%) occurred at the shoulder, 13 (45%) at the elbow, 8 (27%) at the wrist, and 6 (9%) at the hand. We assessed elbow, wrist, and hand EMG activity among agonist and antagonist muscle groups. Agonist muscles are defined as those participating to move the joint in the intended direction and antagonist muscles as those opposing such movements. Figure 3 shows evaluations at which significant changes in agonist and antagonist muscle EMG activity were found for the MC Group. although we would expect increasing values (upward arrows) of EMG from agonist muscles and decreasing values (downward arrows) from antagonist muscles, these relationships were not consistent.

Targeted Training Group. table 5 depicts the 19 functional measures for which significant between-evaluation changes occurred in the TT Group. For the total of 75 significant functional changes across evaluations 5 to 11, 48 (64%) involved shoulder tasks, 11 (15%) were related to the elbow movements, and 16 (21%) involved the wrist and hand tasks. Figure 1 also shows the numbers of significant changes among all tasks (top graph) and composite tasks (lower graphs) for the TT Group. The total number of tasks showing significant improvement was fairly consistent across evaluations until the one-year follow-up evaluation when even more improvement was seen (Fig. 1, top graph). This behavior was mirrored in evaluations 6 to 11 at the shoulder and at the wrist and hands (Fig. 1, lower graphs).

Figure 4 shows the distribution of significant changes in ROM across evaluations 6 to 11 for the TT Group. Among the 20 changes in active ROM, 9 (45%) involved the shoulder, 1 (5%) pertained to the elbow, and 10 (50%) were related to the wrist and hand. There were more consistent total column counts during treatment (evaluations 6-8) than at follow-up (evaluations 9-11).

Table 6 displays statistically significant increases (upward arrows) and decreases (downward arrows) in integrated EMG activity across evaluations 5 to 11 for the TT Group. For the 26 significant changes, 1 (4%) involved the shoulder (anterior deltoid muscle), 14 (54%) occurred at the elbow, 7 (27%) involved muscles acting at the wrist, and 4 (15%) involved muscles acting on hand movements. Figure 5 shows evaluations at which significant changes in agonist and antagonist muscle EMG activity occurred in the TT Group. As with the MC Group, increasing values of EMG activity (upward arrows) for agonist muscles and decreasing values of EMG activity (downward arrows) for antagonist muscles were expected, but these relationships were not consistent in the TT Group.

Between-Groups Treatment

Measurements

Table 7 illustrates where significant differences between groups occurred for the 19 functional measures. Only 6 (24%) of the 25 significant differences found overall between groups occurred in the MC Group, with 4 (67%) occurring at the shoulder and 2 (33%) occurring at the wrist and hand. Nineteen (76%) of the 25 significant differences occured in the TT Group, with 6 (31%) occurring at the shoulder, 10 (53%) at the elbow, and 3 (16%) at the wrist and hand. Figure 6 shows the total number of tasks in evaluations 6 to 11 for which significant between-groups differences could occur as well as those that did occur in the MC and TT Groups. The number of significant changes at each evaluation was fairly consistent in the TT Group, whereas the MC Group demonstrated a majority of significant changes in posttreatment evaluations for tasks involving the shoulder and the wrist and hand.

Of the 16 significant between-groups changes in active ROM, 13 (81%) occurred in the MC Group and 3 (19%) occurred in the TT Group (Fig. 7). The three changes seen in the TT Group were at the wrist. Of the 13 changes in the MC Group, 3 (23%) occurred at the elbow, 2 (15%) involved the wrist, and 8 (62%) occurred at the hand. Significant between-groups ROM changes in the TT Group appeared equally across evaluations 7, 8, and 9. The majority of such changes in the MC Group were at evaluations 10 and 11.

Table 8 presents the significant changes in EMG activity between groups. Of the 22 changes, 14 (64%) were found in the MC Group and 8 (36%) were in the TT Group. In the MC Group, 1 significant change (7%) occurred at the shoulder, 8 (57%) occurred at the elbow, 3 (22%) occurred at the wrist, and 2 (14%) occurred in the fingers. In the TT Group, 3 significant changes (38%) occurred at the elbow, 4 (50%) occurred at the wrist, and 1 (12%) occurred at the hand. Figure 8 illustrates evaluations at which significant changes in agonist and antagonist muscle EMG activity were found between groups. Increasing values of EMG activity (upward arrows) among agonist muscles and decreasing values of EMG activity (downward arrows) among antagonist muscles were not consistent for either group, demonstrating between-groups changes. Table 9 summarizes all within-group and between-groups changes. Table 10 lists a series of tasks that certain patients were unable to complete before treatment but learned to complete after treatment.

Discussion

Results from this study suggest that both conventional biofeedback training (TT), in which all efforts at neuro-muscular reeducation are directed toward the hemiplegic upper extremity, and MC training procedures, in which EMG output from the unaffected upper extremity is used to train its homologous counterpart in the hemiplegic limb, are effective in enhancing motoric behavior among patients with chronic neurologic disorders who possess at least minimal extensor muscle activity. Close scrutiny of data that identify significant changes in functional activities, active ROM, and EMG activity reveals that the time courses over which these within-group changes occur are not the same. Most MC changes occur at follow-up evaluations, suggesting that the training periods may have been a necessary prerequisite for enhanced functional and physiological changes. Furthermore, if MC training represents an effort to make optimal use of bilateral descending motor systems, we would predict primarily enhanced proximal muscle activity because of the predominance of efferent, nonpyramidal synapes upon motoneurons innervating proximal muscles, based on information available from higher primate studies. [12] The Mc Group showed within-group improvements for tasks requiring manipulation of the environment (wrist-hand), active wrist and finger ROM changes, and even distal EMG activity changes, some of which were inappropriate. In addition, several distal functional tasks and active ROM showed significantly greater improvement within the MC Group compared with the TT Group (Tab. 9).

Although a specific explanation for these occurrences is impossible, it should be emphasized that all patients did possess varying degrees of distal muscle activity prior to treatment, and MC applications, much like TT, did provide on-line audiovisual representations of muscle activation that could facilitate appropriate pyramidal drive while also engaging extrapyramidal efferent systems. The latent nature of improvements among MC Group members, although quantitatively similar to the TT Group, suggests that more time and practice may be needed if this approach is to show more benefits earlier in treatment.

Table 10, which lists tasks patients learned to complete after treatment, shows that almost twice as many reacquisitions of function occurred in the TT Group than in the MC Group and may be accounted for, in part, by the fact that the total number of completed tasks in the TT Group was smaller compared with the MC Group. Subtle improvements among some MC Group patients are perplexing because, as seen earlier, most functional and physiological gains occurred at follow-up. The ability of some MC Group patients to complete tasks during treatment indicates that not all significant improvements occurred after treatment.

In examining EMG activity changes, not all significant values were appropriate. For the MC Group, 10 of 11 significant changes in agonist muscle EMG activity were in the appropriate direction (Fig. 3, top diagram), as were 9 of 10 significant changes in agonist muscle EMG activity in the TT Group (Fig. 5, top diagram). Fifteen of 18 antagonist muscle group EMG activity levels were significantly changed in the inappropriate direction for the MC Group (Fig. 3, bottom diagram). Most notably, triceps brachii muscle activity was significantly elevated during elbow flexion for 6 of 6 evaluation comparisons, wrist extensor muscle activity was elevated during wrist flexion for 4 of 5 evaluation comparisons, and biceps brachii muscle activity was significantly augmented during elbow extension for 3 of 5 evaluation comparisons (Fig. 3, bottom diagram). Similarly, in the TT Group, antagonist muscle EMG activity was changed in the inappropriate direction for 13 of 16 significant changes (Fig. 5, bottom diagram), including elevated triceps brachii muscle EMG activity during elbow flexion (7 of 7 evaluation comparisons), elevated wrist extensor muscle EMG activity during wrist flexion (3 of 3 evaluation comparisons), and increased finger extensor muscle EMG activity during finger flexion (2 of 2 evaluation comparisons). Comparing EMG activity changes between groups revealed that agonist muscle changes were usually in the appropriate direction, with significantly greater changes seen in the MC Group on 4 occasions and in the TT Group on 3 occasions (Fig. 8, left diagrams). Antagonist muscle changes between groups, however, showed correct directionality in 4 of 5 occasions for the TT Group but only 3 of 10 times for the MC Group (Fig. 8, right diagrams). Collectively, these data suggest that within each group the newly recruited weaker muscle (usually an extensor muscle) was prone to undergo contraction when patients attempted maximal shortening contraction of the agonist muscle. With respect to between-groups behaviors, this behavior was observed more frequently in the MC Group than in the TT Group. To what extent these individual aberrations in muscle activity affected functional changes cannot be ascertained. Recently, Tyler and Hutton have shown that biceps brachii-triceps brachii muscle contraction ratios vary based on elbow position, muscle length, and direction of movement. [13] These data suggest that a closer scrutiny of agonist-antagonist co-contraction ratios be undertaken to determine whether our treatment strategies actually promote co-contraction, with recruitment of paretic muscles, as a primary vehicle to enhance function. Alternatively, in future studies, a more appropriate manner by which to assess muscle activity among antagonist muscle pairs could be to observe their behaviors during functional activities rather than during isolated joint movement efforts.

Relationship of Functional

Assessments to Physiologic

Measures

To explain why MC or TT techniques can improve movement or function requires seeking correlations between integrated EMG, active ROM, and force- or time-based functional measures. With the exception of a few positive correlations between time to complete simple tasks involving shoulder movements and increased active shoulder ROM, no close relationships were noted, thus confirming our observations for similar measures in a previous study. [11]

The relationship of muscle force to EMG activity under isometric conditions has been reviewed in detail by Stuart and Enoka. [14] Although numerous studies have confirmed linear relationships between these two variables, comparatively limited data are available to describe EMG activity and active movement in human upper extremity musculature during free-moving functional tasks. We have known for some time that a reasonable correlation exists between arm movements and surface and indwelling EMG recordings from muscles producing specific joint actions. [15] Our comprehension of interrelationships between function and EMG activity, however, is confounded by the realization that factors such as muscle length when initiating contraction, [16] velocity of movement, [17] duration of contraction, [18] instructional set relating to the intention of movement, [19] and variability in ease of motor unit recruitment [20] all contribute to the analysis of human movement.

Collectively, these factors appear almost overwhelming to control. When these considerations are added to the hyperactive neuromotor system of the patient with neurologic disorders, it becomes increasingly clear that the complexity of interrelated issues transcends what have been traditional, simplistic measures to assess unimpeded limb movement: functional activities, quantified EMG activity, and ROM. Perhaps more emphasis should be placed on kinematic analyses that bypass some of the difficulties inherent within these physiologic measures. To maximally control for the important variables noted above may necessitate movement paradigms that are not truly functional, that is, fixation of a limb segment or the superimposition of specified loads. This paradox is complicated further by the knowledge that changing central motor planning to achieve a movement goal may alter the quality, quantity, and timing of muscle activity. Consequently, conclusions about mechanisms to account for the efficacy of training techniques such as MC or conventional EMG biofeedback must be guarded at best.

Conclusions

Both the TT and MC techniques promote enhanced function among selected patients with neurologic disorders. Clients receiving conventional EMG biofeedback training (TT) appear to produce significant changes during treatment. Although the comparable but more latent benefits of MC traning provide less convincing evidence for direct treatment efficacy, much of the training strategy provides self-treatment because patients match muscle outputs to the specifications (individual amplifier gain controls) set by the therapist. In this regard, the procedure becomes more cost-effective because the therapist can devote time to other patients as individual MC self-treatment takes place. Equally important, however, is the realization that the data acquisition in this study is comparable to documentation possibilities in the clinic (eg, ROM, EMG, values, timed functional tasks) and in this regard may serve as a model by which clinicians can seek to quantitatively efficacy of intervention procedures for patients with neurologic disorders. With respect to specific MC or TT approaches, our analyses suggest that these approaches can be applied to chronic left- or right-hemiplegic or head-injured patients, all of whom have some upper extremity extensor muscle function.

Future efforts should be directed toward replicating our results as well as determining the relative effectiveness of each treatment approach in more acutely injured patients with neurologic disorders among whom more dynamic neuroplastic events presumably occur. Whether seeking a clearer explanation for the modus operandi of these treatment techniques or for other approaches to clients with movement impairments, therapists also must recognize the limited value of quantified EMG activity to explain major functional changes. More sophisticated or alternative measures are needed to allow reliable assessment of such demonstrably important factors as muscle length, velocity of movement, functional relevance of movement, and motoric strategies necessitating co-contraction or reciprocal inhibition of antagonistic muscle groups.

Acknowledgments

The software packages for EMG analyses, graphic displays, and data formatting were developed by Jim Hudson, whose invaluable assistance is much appreciated. We thank Silvia Sullivan for her help with the illustrations, Vera Backstrom and Ronda Parkison for their flawless typing skills, and George Cotsonis for his consultation on statistical analyses.

(*1) International Business Machines Corp, Old Orchard Rd, Armonk, NY 10504.

(*2) Hyperion, Inc, 9579 SW 168th St, Miami, FL 33157.

References

[1] Ghez C: Introduction to the motor systems. In Kandel ER, Schwartz JH (eds): Principles of Neural Science. Amsterdam, The Netherlands, Elsevier Science Publishers BV, 1985, pp 436-440

[2] Brooks VB: The Neural Basis of Motor Control. New York, NY, Oxford University Press Inc, 1986, pp 93-104

[3] Kuypers HGJM: Central cortical projections to motr and somato-sensory cell groups. Brain 83:161-184, 1960

[4] Liu CN, Chambers WW: An experimental study of the corticospinal system in the monkey (Macaca mulatta). J. Comp Neurol 123:257-283, 1964

[5] Phillips CG, Porter R: The pyramidal projection to motoneurons of some muscle groups of the baboon's forelimb. Prog Brain Res 12:222-242, 1964

[6] Brinkman J, Kuypers HGJM: Cerebral control of contralateral and ipsilateral arm, hand and finger movements in the split-brain rhesus monkey. Brain 95:653-674, 1973

[7] Matsunami K, Hamada I: Characteristics of theipsilateral movement-related neuron in the motor cortex of the monkey. Brain Res 205:29-42, 1981

[8] Booker HE, Rubow RT, Coleman PJ: Simplified feedback in neuromuscular retraining: An automated approach using electromyographic signals. Arch Phys Med Rehabil 49:621-625, 1969

[9] Kelly JL, Baker MP, Wolf SL: Procedures for EMG biofeedback training in involved extremities of hemiplegic patients. Phys Ther 59:1500-1507, 1979

[10] Wolf SL, Edwards DI: Comparing targeted feedback and motor copy training procedures in the restoration of upper extremity function among chronic neurologic patients. Abstracts of the Society for Neuroscience 13:1700, 1987

[11] Wolf SL, Binder-Macleod SA: Electromyographic biofeedback applications to the hemiplegic patient: Changes in upper extremity neuromuscular and functional status. Phys Ther 63:1393-1403, 1983

[12] Humphrey DR: On the cortical control of visually directed reaching: Contributions of nonprecentral motor area. In Talbott RE, Humphrey DR (eds): Posture and Movement. New York, NY, Raven Press, 1979, pp 51-112

[13] Tyler AE, Hutton RS: Was Sherrington right about co-contractions? Brain Res 370:171-175, 1986

[14] Stuart DG, Enoka RM: Motoneurons, motor units, and the size principle. In Rosenberg RN (ed): The Clinical Neurosciences. New York, NY, Churchill Livingstone Inc, 1983, pp 471-517

[15] Bouisset S, Maton B: Quantitative relationship between surface EMG and intramuscular electromyographic activity in voluntary movement. Am J Phys Med 51:285-295, 1972

[16] Vredenbregt J, Rau G: Surface electromyography in relation to force, muscle length and endurance. In Desmedt JE (ed): New Developments in Electromyography and Clinical Neurophysiology. Basel, Switzerland, S Karger AG, Medical and Scientific Publishers, 1973, vol 1, pp 607-622

[17] Matom B, LeBozec S, Crockaert JC: The synergy of elbow extensor muscles during dynamic work in man: II. Braking of elbow flexion. Eur J Appl Physiol 44:271-278, 1980

[18] Stulen FB, DeLuca CJ: The relation between the myoelectric signal and physiological properties of constant-force isometric contractions. Electroencephalogr Clin Neurophysiol 45:681-698, 1978

[19] Brown SHC, Cooke, JD: Amplitude- and instruction-dependent modulation of movement-related electromyogram activity in humans. J Physiol (Lond) 316:97-107, 1981

[20] Belanger AY, McComas AJ: Extent of motor unit activation during effort. J Appl Physiol: Respirat Environ Exercise Physiol 51:1131-1135, 1981

S Wolf, PhD, FAPTA, is Professor, Department of Rehabilitation Medicine, and Associate Professor, Deparments of Anatomy and Cell Biology and Surgery, Emory University School of Medicine, 1441 Clifton Rd NE, Atlanta, GA 30322 (USA).

De LeCraw, MMSc, is Senior Research Physical Therapist, Emory University Rehabilitation Research and Training Center, Emory University School of Medicine.

L Barton, MMSc, is Supervising Therapist, Physical Therapy Department, Donald Sharp Memorial Hospital, 7901 Frost St, San Diego, CA 92123.

This study was supported in part by Grant No. G00830041 from the National Institute on disability and Rehabilitation Research, US Department of Education, Washington, DC.

This article was submitted June 13, 1988; was with the authors for revision for 17 weeks; and was accepted March 23, 1989.