Agonist and antagonist activity during voluntary upper-limb movement in patients with stroke.

One of the most devastating and common consequences of stroke is loss of the use of the arm and hand. Only about 5% of individuals paralyzed by stroke regain full arm and hand function, and about 20% regain no functional use.[1] The prognosis for upper-limb recovery following stroke remains dismal in spite of intensive therapeutic efforts.[1] This unfortunate state calls for renewed efforts to examine the theoretical basis of the factors impeding upper-limb recovery and their clinical implications.

The basis of effective therapy is valid theories and sound theoretical models.[2] Often, however, differing theories are developed, shaped, and championed. The theoretical rationale underlying popular treatment approaches may become outdated and the clinical practice model not adjusted as new knowledge is gained. One example of how differing theories have developed relates to the treatment of the upper limb. It is the fundamental disagreement about the importance of preventing abnormal degrees of muscle co-contraction while denying the importance of muscle weakness as a major cause of upper-limb dysfunction. At least one proponent of treatment encourages considering both muscle weakness and co-contraction,[3] whereas Bobath,[4,5] whose theories and techniques still form the basis of the widely used neurodevelopmental therapy (NDT), suggests systematically avoiding the use of strengthening techniques. According to Bobath, the "patient's problem is not a lack of muscle power on the affected side."[4](p59) She explained that the weakness of muscles may be due to "exaggerated co-contraction, a typical feature of spasticity."[5] Bobath theorized that the exaggerated co-contraction does not give the spastic agonists a chance to contract against the resistance of equally strong, or stronger, antagonists.[5] She warned therapists about the use of excessive effort, which is believed to lead to an increase of co-contraction, spasticity, and widespread associated reactions.[4](pp 18,19) Because Bobath believed that the problem was not a lack of muscle power, an important tenet or principle of this treatment approach is to avoid having patients expend the effort involved in strengthening muscles.[4](p59) Although Bobath-based NDT is still favored by many therapists working with patients who have strokes, the evidence on which to base a decision to support or refute this particular Bobath tenet has not been clear.

Literature Review

The research literature discusses both muscle weakness attributable to inadequate motor unit recruitment and co-contraction attributable to impaired antagonist inhibition.

Muscle Weakness Attributable to Inadequate Motor Unit Recruitment

Almost from the start, evidence challenged Bobath's tenet that muscle weakness is not a problem in upper-limb dysfunction; inadequate recruitment of motor units, which would result in an inability to generate sufficient force, was demonstrated to be an important reason for poor motor performance in patients with stroke.[6-12] Both alterations in the frequency of firing of motor units[7] and a reduction in the number of active motor units[12-14] were offered as explanations for the reduced recruitment. Additionally, the loss of orderly recruitment and rate modulation of motoneurons within a given motoneuron pool were shown to lead to inefficient muscle activation, inducing early loss of force, increased effort, and the clinical perception of weakness.[15] Burke[16] describes features of the upper motoneuron syndrome as both "negative" and "positive." Negative features include weakness and loss of dexterity, particularly fine manual manipulation, and positive features include abnormal posture, spasticity, and exaggeration of some exteroceptive (cutaneous) reflexes. Burke points out that, for the majority of patients with upper motoneuron syndrome, the major defects in function are negative, not positive. The importance of negative symptoms is not surprising, given that the pyramidal tract is the executive pathway for volitional goal-directed movement. Any interruption of that pathway produces a deficit in those muscles that normally act as prime movers in volitional movement.[6,15,16]

Co-contraction Attributable to Impaired Antagonist Inhibition

Co-contraction of agonist and antagonist muscles attributable to impaired antagonist inhibition on the hemiplegic side is a recognized clinical phenomenon in persons who have had a cerebrovascular accident.[17] There is controversy, however, about the importance and prevalence of this "abnormal" co-contraction. Several authors[9,17,18] found abnormal co-contraction to be a limiting factor in the movement of patients with stroke. Knuttson and Martensson[9] summarized these findings by concluding that the abnormal muscle co-contraction or spastic restraint that was seen particularly during voluntary motion may constitute a crucial component in the motor handicap. Two different types of studies arrived at these conclusions. The first type of study used a comparison of the antagonist versus agonist functions of one muscle during opposite motions.[9,18] The second type of study investigated the simultaneous electromyographic (EMG) activity of agonist and antagonist muscles.[17]

Using different methods with simultaneous agonist-antagonist EMG recordings in a study of voluntary muscle strength at the elbow, Colebatch et al,[12] with a more sophisticated technique involving torque measures, found no evidence of co-contraction in 17 out of 18 patients. Similar studies have found little evidence of abnormal co-contraction at the ankle,[8] the elbow,[10] and the upper limb.[11] One author[19] reports that the EMG amplitudes in both the agonists and the antagonists of patients with stroke were decreased. These conflicting findings are best summarized by Hammond et al,[17] who suggest that the opposing views may be due to the manner in which co-contraction is considered and analyzed. We agree with this suggestion and propose that any method of measuring co-contraction in the presence of inadequate recruitment should take this inadequate recruitment into consideration. A method using an agonist: antagonist ratio would skew the results to erroneously indicate abnormally high degrees of co-contraction. In summary, the controversy is whether activity in the antagonist muscles is sufficiently excessive to block movement or only high when considered relative to the inadequate recruitment of the agonists. A solution to this controversy would be important if the answer suggests a revised approach to the treatment of the patient who has a hemiplegic upper limb.

Against this background, we designed and carried out a study to examine the movement disorder in the hemiplegic upper limb. The objective was to determine the relative contributions of altered motor unit recuritment and altered activity of the antagonist muscles during attempts to complete movement tasks of varying complexity. Our observations in the clinical use of EMG biofeedback had led us to question the hypothesis that movement is blocked by co-contraction attributable to impaired antagonist inhibition.

Method

Subjects

This quantitative laboratory study included 44 patients with hemiplegia resulting from a recent stroke. The median number of weeks post-onset was 11 (range=3-38). The patients were 20 women and 24 men, with a mean age of 67 years (range=33-85). All patients had unilateral involvement, with the hemiplegia on the right side in 21 patients and on the left side in 23 patients. Ten age-matched, nondisabled volunteers (5 men, 5 women), with a mean age of 64 years (range=39-78), were recruited from the community to serve as a control group.

The Chedoke-McMaster Stroke Assessment[20-22] was used to determine the stage of motor recovery of the arm of all subjects. The stages described in this measure are based on those originally described by Brunnstrom.[23] The Chedoke-McMaster Stroke Assessment has been shown to yield both valid and reliable results.[20,21] Detailed guidelines include a description of the tasks at each stage, the patient's starting position, the therapist's instructions to the patient, and what constitutes acceptable and unacceptable performance.[24] Motor recovery in the involved arm ranged from stage 2 to 6, with 10 patients in each of stages 2, 3, and 6 and 7 patients in each of stages 4 and 5. The control subjects all scored in stage 7 (ie, normal). All subjects provided written informed consent.

Measures

Temporal, spatial, and EMG data were collected from all subjects during the accomplishment or the attempted accomplishment of six voluntary movement tasks. The EMG data were gathered from the superficial proximal arm and shoulder musculature.

Movement tasks. The six tasks were chosen from the Chedoke-McMaster Stroke Assessment.[20] Tasks of increasing complexity involving a variety of movements and upper-limb muscles were selected from the stages of motor recovery to represent a spectrum of activities. The tasks (along with the corresponding stage of recovery) were as follows: (1) touch opposite knee (stage 3), (2) hand to chin (stage 3), (3) flexion synergy(*1) (stage 4), (4) shoulder flexion to 90 degrees with elbow extension (stage 4), (5) shoulder abduction to 90 degrees with elbow extension (stage 5), and (6) raise arm overhead with elbow extension (stage 6). All patients attempted all tasks. Data were analyzed from both patients who succeeded and those who did not succeed in completing a task. The number of tasks that any one patient could succeed in completing depended on that patient's stage of recovery (ie, patients with the arm in stage 2 could complete no tasks, those with the arm in stage 3 could complete only the two stage-3 tasks, those with the arm in stage 4 could complete only the four stage-3 and -4 tasks, and so forth). Patients who could not succeed in completing a movement task were asked to attempt as much movement as possible in the allotted time period.

Baseline EMG values, with the limb at rest, were obtained for each subject before the tasks were carried out in order to identify the signal level attributable to ambient instrumentation noise and isoelectric EMG activity. Prior to data collection for each task, patients practiced to ensure that their performance was typical of their ability. A 30-second rest period was allowed before and after data collection for each task. Within a 5-second period, the subject was required to perform, or attempt to perform, the movement and return to the resting position according to the standardized protocol. The six tasks were carried out in the order described earlier.

Muscle selection. Nine proximal and relatively superficial upper-limb muscles (ie, upper trapezius; rhomboids; clavicular and sternal heads of the pectoralis major; anterior, middle, and posterior deltoids; biceps brachii; and triceps brachii) were monitored using surface EMG.

Instrumentation. The EMG electrodes were applied to both upper limbs using a specified protocol. This protocol was developed in our laboratory in order to prepare the text entitled Electrode Placement in EMG Biofeedback.[25] This text describes the technique used for skin preparation and surface electrode selection, preparation, placement, and attachment. This protocol was designed to minimize cross talk and ensure an acceptable level of electrode impedance. Pairs of miniature surface electrodes[unkeyable] (3-mm diameter) were placed with minimal interelectrode distance on the identified muscles as described and diagrammed in the text. A Selspot movement analysis system[unkeyable] was used to track the muscles' performance. This information was needed to identify the initiation of movement and arrival at the target (Figs. 1 and 2). Six infrared light-emitting diodes were attached at the knuckles (1 each over the dorsum of the heads of the first and fifth metacarpals), wrist (1 each over the styloid processes of the radius and the ulna), shoulder (over the greater tuberosity of the humerus), and target. The pairs of diodes at the wrist and knuckles served to ensure that the signal did not move out of camera view during movement. Successful completion of the task was confirmed by the arrival of either of the distal diodes at the target diode. The target diode was repositioned for each task.

Data Collection and Processing

Following the command to start each of the seven trials (one resting and six tasks per side), EMG and Selspot movement data were collected on-line for 7.5 seconds at a sampling rate of 80 Hz per channel by a PDP 11/34 computer[unkeyable] equipped with a 16-channel, 12-bit analog-to-digital converter. The EMG signals were first differentially amplified, then bandpass filtered from 40 to 500 Hz and processed (full-wave rectified and 25-Hz low-pass filtered) by the analog signal processing hardware. The data were further smoothed using a 5-Hz zero-phase-shift digital low-pass filter. Raw EMG signals were inspected on an oscilloscope, and graphic hard-copy displays of the EMG envelopes and movement data were generated and visually inspected to monitor the artifacts of the data. Trials were immediately repeated when problems such as instrumentation failure, motion artifact, or atypical performance were observed.

Data Analysis

The mean amplitude obtained during 6 seconds of resting EMG activity was used as the baseline value. For each trial, data were visually inspected to determine the phase from initiation of movement to arrival at the target. For patients who were unable to reach the target, the phase was considered complete when the maximum excursion of the limb was reached. Mean EMG amplitudes (and standard deviations) were calculated for this phase by first subtracting the baseline value, then taking the area under the curve (the EMG envelope) and dividing it by the time duration of the contraction in order to normalize the value with respect to time. To identify agonists and antagonists for each of the six tasks, the average EMG data from the 10 control subjects (20 sides) were used. There were three steps in identifying these muscles. In step 1, for each subject, a 9 x 6 table was constructed for each side, showing the average EMG values for each muscle during each task. Each muscle's values were then examined to determine the task or tasks during which it was most active. For example, the posterior deltoid muscle was often most active during shoulder abduction to 90 degrees. For each task, identified muscles were flagged if the activity level was substantially above baseline (greater than 20 [unkeyable]V), which in our experience eliminates the inclusion of random low-level muscle activity. Where more than one muscle contributed in a similar fashion to a task, all were flagged. In step 2, the findings from the 10 control subjects were combined. If, for a given task, a muscle was flagged in more than 25% of the subjects (set arbitrarily), the muscle was classified as an agonist for that task. In step 3, muscles that are known to have actions primarily in the opposite direction were identified as antagonists.[26]

Normalization 

Since the absolute EMG amplitude of muscle activity in different muscles is influenced by variable distances of recording electrodes from active tissues, it cannot be used for direct comparisons.[6] Instead, some form of normalization was required. In nondisabled subjects, it is the ratio of the EMG amplitude to the maximal voluntary contraction value for the same muscle that is used. Following stroke, however, when the maximal voluntary contraction is a meaningless value because of paresis ("attenuation of electrogenesis during voluntary contraction"[27](p591), an alternate means of quantifying EMG amplitudes must be used. Knutsson and Martensson[9] used the average amplitudes. Several authors[10,12,17,27] performed a cross-muscle analysis and calculated a ratio of the agonist to the antagonist. Sahrmann and Norton[6] suggested some form of ratio but opted to use the ratio of the EMG value from voluntary stretch versus that from passive stretch in the same muscle. They then compared values from the involved limb with those of the contralateral, uninvolved side.

We felt that none of these forms of depicting EMG values allow for an accurate representation of the activity in the agonist. Instead, we opted to compare the values on the hemiplegic side with those of the same muscle on the uninvolved side (ratio of involved side to uninvolved side). This approach was used by Tang and Rymer and "... is based on the premise that patterns of motor unit activation in the non-paretic limb of the subject with hemiplegia are essentially normal."[10](p694) In addition, for patients with stroke in the postacute phase, the degree of atrophy is minor and the geometry of the recording electrodes versus active tissue is essentially unchanged. Although the standard normalization procedure would have been feasible for our control subjects, the data obtained could not have been compared with that of the patients; therefore, in the control subjects, the ratio of the same muscle from the right to left sides was used. A logarithmic transformation of the ratios was carried out to linearize the sensitivity.[28] Otherwise, the range of values for the denominator larger than or equal to the numerator would be 0 to 1, and the numerator larger than or equal to the denominator will result in ratios of 1 to infinity. [TABULAR DATA OMITTED]

Statistical Analyses

All statistical analyses were performed using SPSS/PC + Version 3.1 software.[unkeyable] Using the log of the ratio of the involved and uninvolved limbs, unpaired t tests were performed to determine significant differences between those patients who succeeded and those who did not succeed in completing the tasks.

As we were dealing with several dependent variables for each task--the groups of agonist and antagonist muscles for each--a multivariate analysis of variance (MANOVA) was carried out on each of these sets of variables in order to examine overall differences between the two groups of patients. Again, the log of the ratio of the involved limb to the uninvolved limb was used.

Results

The results of the exercise that was undertaken to identify agonists and antagonists for each of the six tasks from the 10 control subjects are presented in Table 1. The muscles are given in order of frequency of occurrence among these subjects.

Figures 1 and 2 provide a graphical depiction of the temporal, spatial, and EMG data from two patients with hemiplegia. Over a 5-second time frame, the proximal and distal markers display the spatial excursion and the EMG recordings depict the muscle activity. Figure 1 displays the data from the involved limb of a typical patient with hemiplegia who, with stage-2 recovery, could not complete the task of flexing the shoulder to 90 degrees with the elbow extended. By definition, at this stage of recovery, voluntary movement does not occur. In this figure, the unshaded portion depicts a slight commencement of movement at the shoulder. Little or no muscle activity was observed in both the agonists (anterior deltoid, triceps brachii, upper trapezius muscles) and the antagonists (posterior deltoid, biceps brachii muscles) during the patient's attempt to move. Only a few microvolts of activity above threshold were recorded in the upper trapezius and biceps brachii muscles. This activity remained relatively constant throughout the 5-second time frame and was inadequate to bring about movement. A small, but inadequate, burst of appropriate activity occurred in the triceps brachii muscle as the patient attempted to flex the shoulder.

Figure 2 shows the results from a typical patient with hemiplegia who, with stage-5 recovery, could complete the same task of flexing the shoulder forward to 90 degrees. The most striking feature is the proportional increase in EMG activity in almost all muscles when compared with the patient with stage-2 recovery. In this patient, just before movement commenced (just before the unshaded column starts), a rapid increase in activity occurred in several muscles, particularly the anterior and middle deltoid muscles. There was also a short burst of activity in the triceps brachii muscle once the patient's arm started forward. Although considerable activity was seen in the antagonists (posterior deltoid and biceps brachii muscles), they did not block movement--this patient succeeded in completing the task.

[TABULAR DATA OMITTED]

Table 2 shows the means and standard deviations of the agonists on the involved and uninvolved limbs for the patients who completed versus those who did not complete the six movement tasks. The mean values show the direction of any differences between the groups. The results of the t tests (performed on the log of the ratio of the involved and uninvolved limbs) were used to compare these two groups. In over half of the muscles (13 of 23), significant differences between the two groups of patients were found. When the groups varied significantly, the average EMG values from the involved limb were always lower in the patients who were unable to complete the task. The results of the MANOVA and significance levels, for each set of variables, were used to examine overall differences between the groups. For all six tasks, the EMG activity of the set of muscles in the patients who were unable to perform the tasks was significantly lower than in those who could perform the tasks.

[TABULAR DATA OMITTED]

Table 3 shows the results from the antagonist muscles. The analyses were similar to those reported in Table 2. In comparison with the data from the agonist muscles, the data from the antagonists did not differ markedly between the two groups. On only one occasion was a significant difference between the patients who could and those who could not perform the task noted. In this one instance (ie, biceps brachii muscle for the task "touch opposite knee"), the values were again lower for the group who could not perform the task. There was no evidence of a significant increase in antagonistic activity in the patients who were unable to complete the tasks. Similar results were found when data from the sets of muscles were analyzed.

In a manner similar to the comparisons of the patients who completed and those who did not complete the specified tasks, statistical comparisons were made between the patients who were able to complete the tasks and the group of control subjects (tables not included). No statistically significant differences were noted between these two groups; that is, for all six tasks, when the groups of muscles were examined by MANOVA, no significant differences were found.

In summary, for all tasks, patients who did not complete the tasks showed significantly less agonist muscle activity than those who did complete the tasks. In muscles acting as antagonists, a significant difference overall was seen only for the task "touch opposite knee."

Discussion

Our data clearly show that the significant factor separating the group of patients who were unable to complete an arm movement task from the group who did complete the task was inadequate recruitment in the agonist muscles, not abnormal co-contraction of the agonist and antagonist muscles. We found no evidence of abnormally high EMG activity in the upper-limb antagonist muscles of patients with paresis. This finding of inadequate motor unit recruitment is in general agreement with the findings of authors[6-12] who associate inadequate recruitment of motor units with an inability to generate sufficient force to accomplish motor tasks.

With a significant reduction in agonist activity, it is not surprising that studies involving a co-contraction ratio of agonist to antagonist would show excessive co-contraction of muscles, as has been the case.[9,17] Bobath[4,5] appears to have surmised incorrectly that excessive activity in the antagonists, and not inadequate motor unit recruitment causing weakness in the agonists, was the major cause of an inability to move. As Duncan and Badke point out:

Clinicians should realize that in stroke rehabilitation it is more appropriate to concentrate upon reestablishing normal active motor control rather than reducing the hypersensitivity of the stretch reflex in response to passive movement.[3](p147)

The design of this study allowed us to systematically examine differences between groups of patients who could and could not complete specific movement tasks. The significant differences between these two groups suggest that this is a logical way of separating patients for study purposes. Also, by comparing a group of patients who could complete the tasks with a matched group of nondisabled subjects and showing that they did not differ significantly, we have demonstrated that it is the degree of paresis, not the presence or absence of the diagnosis of stroke or hemiplegia, that is the critical factor in explaining the loss of function in this population.

It should be noted that some subjects shifted from the group who completed a task to the group who did not complete the task, depending on the complexity of the task. For example, 34 patients could complete tasks 1 and 2, whereas only 10 patients could complete task 6. Again, it is the presence or absence of ability to perform the task that explains the results. These findings suggest that an individual can vary his or her performance, depending on the complexity of the task. These results are consistent with the theoretical construct underlying the stages of recovery as measured by Brunnstrom[23] and Fugl-Meyer et al.[29]

Although the clinical assessment used has been tested for reliability and recently reported,[20,21] the reliability of our laboratory data has not been established. Addressing this issue was not feasible, because the patients with paresis were unable to complete multiple trials at one time without a change in performance because of fatigue, a known problem in this population. Instead, all patients practiced the movements only long enough to ensure that the performance was typical. The advantage of completing the study as designed was that the data from an individual subject could be collected during one session. Variability attributable to a change in electrode placement was avoided, and an optimal amount of information was obtained. The issue of the number of trials needed to collect data yielding reliable results from the upper limb requires further exploration. In studies of gait that have demonstrated a high level of stability, the number of strides recommended to provide reliable information is three.[30] Because movements in the arm involve more degrees of freedom and are generally more variable than for the leg, the number of trials needed would probably be greater. Future work is needed to determine the number of trials required to minimize variability of the EMG data in voluntary arm movements. From a clinical perspective, this study adds to the body of knowledge that supports aiming treatment efforts at improving motoneuron recruitment and increasing strength in the agonist muscles.[31,32]

Conclusions

The evidence from this study suggests that treatments aimed at reducing abnormal co-contraction as part of the overall management of the upper limb of patients with hemiplegia following stroke may not be well founded. Bobath's tenet[4,5] that abnormal degrees of co-contraction are what block movement may actually be invalid in this population. The efficacy of treatment aimed at increasing motor unit recruitment requires careful study. This is not to suggest that the patients who were unable to perform the movements did not have spasticity. Rather, the findings of this study support Burke's statement that in "the majority of patients with the upper motoneuron syndrome, whether they have developed spasticity or not, the major defects in function are negative, not positive."16(p402) Cerebral shock, weakness, and loss of dexterity are greater problems than are spasticity and the resistance to movement attributable to co-contraction of the agonist and antagonist muscles.

(*1)Flexion synergy includes shoulder horizontal abduction to 90 degrees, with sufficient shoulder medial rotation, elbow flexion, and forearm supination for the hand to touch the ear.

[unkeyable]Model #650414, Beckman Instruments Inc, 3900 River Rd, Schiller Park, IL 60176.

[unkeyable]Selective Electronics Co, AB Partille, Sweden.

[unkeyable]SPSS Inc, 444 N Michigan Ave, Chicago, IL 60611.

References

1 Basmajian JV, Gowland C, Brandstater M, et al. EMG feedback treatment of upper limb in hemiplegic stroke patients: a pilot study. Arch Phys Med Rehabil. 1982;63:613-615.

2 Harris SR. Therapeutic exercises for children with neurodevelopmental disabilities. In: Basmajian JV, Wolf SL, eds. Therapeutic Exercise. Baltimore, Md: Williams & Wilkins; 1990:163-176.

3 Duncan PW, Badke M. Stroke Rehabilitation: The Recovery of Motor Control. Chicago, Ill: Year Book Medical Publishers Inc; 1987.

4 Bobath B. Adult Hemiplegia: Evaluation and Treatment. London, England: William Heinemann Medical Books Ltd; 1978.

5 Bobath B. Treatment of adult hemiplegia. Physiotherapy Canada. 1977;63:310-313.

6 Sahrmann SA, Norton BJ. The relationship of voluntary movement to spasticity in the upper motor neuron syndrome. Ann Neurol. 1977;2:460-465.

7 Knutsson E, Richards C. Different types of disturbed motor control in gait of hemiparetic patients. Brain. 1979;102:405-430.

8 Rosenfalck A, Andreassen S. Impaired regulation of force and firing pattern of single motor units in patients with spasticity. J Neurol Neurosurg Psychiatry. 1980;43:907-916.

9 Knutsson E, Martensson C. Dynamic motor capacity in spastic paresis and its relationship to prime mover dysfunction, spastic reflexes and antagonistic co-ordination. Scand J Rehabil Med. 1980;12:93-106.

10 Tang A, Rymer WZ. Abnormal force: EMG relations in paretic limbs of hemiparetic human subjects. J Neurol Neurosurg Psychiatry. 1981;44:690-698.

11 Whitley DL, Sahrmann SA, Norton BJ. Patterns of muscle activity in the hemiplegic upper extremity. Phys Ther. 1982;62:641. Abstract.

12 Colebatch JG, Gandevia SC, Spira PJ. Voluntary muscle strength in hemiparesis: distribution of weakness at the elbow. J Neurol Neurosurg Psychiatry. 1986;49:1019-1024.

13 McComas AJ, Sica RE, Upton ARM, Aguilera N. Functional changes in motoneurones of hemiparetic patients. J Neurol Neurosurg Psychiatry. 1973;36:183-193.

14 Spaans F, Wilts G. Denervation due to lesions of the central nervous system: an EMG study in cases of cerebral contusion and cerebrovascular accidents. J Neurol Sci. 1982;57:291-305.

15 Katz RT, Rymer WZ. Spastic hypertonia: mechanisms and measurement. Arch Phys Med Rehabil. 1989;70:144-154.

16 Burke D. Spasticity as an adaptation to pyramidal tract injury. Adv Neurol. 1988;47:401-422.

17 Hammond MC, Fitts SS, Kraft GH, et al. Co-contraction in the hemiparetic forearm: quantitative EMG evaluation. Arch Phys Med Rehabil. 1988;69:348-351.

18 Yusevich YS. Significance of "global" electromyography for analysing pathological mechanisms of spastic paralysis. Electromyography. 1968;8:135-157.

19 Di Fabio RP. Lower extremity antagonist muscle response following standing perturbation in subjects with cerebrovascular disease. Brain Res. 1987;406:43-51.

20 Gowland C. Standardized physical therapy measurements for assessing impairment and disability following stroke. Neurology Report. 1991;15:9-14.

21 Gowland C, Torresin W, Stratford P, et al. Chedoke-McMaster Stroke Assessment: a comprehensive clinical and research measure. In: Proceedings of World Confederation for Physical Therapy 11th International Congress. 1991:851-853.

22 Gowland C. Staging motor impairment after stroke. Stroke. 1990;21(suppl):I19-II21.

23 Brunnstrom S. Movement Therapy in Hemiplegia: A Neurophysiological Approach. New York, NY: Harper & Row, Publishers Inc; 1970.

24 Chedoke Rehabilitation Centre Stroke Team. Chedoke-McMaster Stroke Assessment: Guidelines and Scoresheet. Hamilton, Ontario, Canada: Chedoke-McMaster Hospital; 1989.

25 Basmajian JV, Blumenstein R. Electrode Placement in EMG Biofeedback. Baltimore, Md: Williams & Wilkins; 1980.

26 Basmajian JV, DeLuca C. Muscles Alive: Their Functions Revealed by Electromyography. 5th ed. Baltimore, Md: Williams & Wilkins; 1986.

27 Visser SL, Aanen A. Evaluation of EMG parameters for analysis and quantification of hemiparesis. Electroencephalogr Clin Neurophysiol. 1981;21:591-610.

28 Zar JH. Biostatistical Analysis. Englewood Cliffs, NJ: Prentice-Hall; 1974:184.

29 Fugl-Meyer AR, Jaasko L, Leyman I, et al. The post-stroke hemiplegic patient, I: a method for evaluation of physical performance. Scand J Rehabil Med. 1975;7:13-31.

30 Arsenault AB, Winter DA, Marteniuk RG, Hayes KC. How many strides are required for the analysis of electromyographic data in gait? Scand J Rehabil Med. 1986;18:133-135.

31 Bohannon RW, Smith MB. Upper extremity strength deficits in hemiplegic stroke patients: relationship between admission and discharge assessment and time since onset. Arch Phys Med Rehabil. 1987;68:155-128.

32 Bourbonnais D, Vanden Noven S. Weakness in patients with hemiparesis. Am J Occup Ther. 1989;43:313-319.