Stroke is the leading cause of long-term disability worldwide1; therefore, assisting an individual's adaptation to deficits can improve function and participation. Reduced upper limb function impacts the ability to perform activities of daily living,2 reducing an individual's independence, and increasing the burden of care for caregivers. The overall cost of health care is increased when individuals with disability are unable to care for themselves. Therefore, the development and refinement of rehabilitation strategies for individuals poststroke have the potential to improve an individual's function as well as decrease the burden on caregivers and on the health care system.
One goal of rehabilitation after stroke is to enhance motor recovery of the paretic upper limb while minimizing compensatory strategies and avoiding learned nonuse.3 Task-related training (TRT), practice of goal-directed functional movements in a natural environment,4 is a common rehabilitation approach that addresses these goals. Task-related training involves variable practice to help the individual develop optimal control strategies for solving motor problems.5 Studies have demonstrated that following a stroke, TRT enhances recovery of locomotion and lower-limb weight bearing in sitting and standing up.6,7 Furthermore, using everyday objects has been shown to improve reaching with the paretic arm during a single test session.8
Restricting trunk motion in individuals poststroke has been shown to encourage more typical elbow and shoulder motion during reaching-to-grasp of objects.9 However, the decreased trunk flexion, increased elbow extension, and improved elbow/shoulder coordination observed during reaching with the trunk unrestrained were tested only immediately following a single training session (60 trials) with a trunk restraint. In a training study of individuals with arm impairment following mild stroke, Michaelson et al10 compared outcomes in a group trained with TRT in conjunction with trunk restraint to a control group trained without trunk restraint; TRT with trunk restraint was associated with greater improvements in function.
Thielman et al3 compared the effects of TRT with those of resistive exercise (RE) on reaching with the paretic arm to targets placed within arm's length after extended practice (12 sessions over 4 weeks, 180 movements/session). While recovery of upper extremity function has typically been poor for individuals with more severe impairment of arm movement,11 in this study, the authors found greater improvement after training with TRT than after RE, as evidenced by straighter hand paths and improved performance on the Rivermead Motor Assessment Scale.12 However, after TRT, compensatory trunk motion (anterior trunk displacement) increased when reaching toward the difficult ipsilateral target. As has been shown previously,13 individuals with more impaired arm movement recruited the additional degree of freedom provided by trunk motion to compensate for inadequate control of the paretic arm. When subjects in the study by Thielman et al3 were tested 1 year after training, they continued to use this compensatory strategy.14
Extending these prior findings, Thielman et al15 compared outcomes associated with extensive training with TRT with those associated with extensive training with RE, both with trunk motion restrained, on reaching with the paretic arm in individuals with moderate or severe motor impairment. Rather than using a control group receiving no therapy, RE was used as the control condition. Results were consistent with prior studies, as kinematic analysis indicated that reaching improved to a greater extent after TRT, as evidenced by straighter hand paths, decreased proportion of time after first velocity peak, and increased upper arm flexion at the difficult ipsilateral target.3 Both training protocols were associated with decreased compensatory trunk flexion, a shift in scapular motion toward protraction, and increased elbow extension.15 In addition, after both TRT and RE, active range of motion of the elbow and performance on the Fugl-Meyer Scale (FMA) improved.16 Thus, restraining trunk motion during extended reaching practice generally led to improvements. These prior findings raise the question of whether individuals with moderate or severe arm impairment can be trained not to use trunk motion without the constant feedback of a restraint.
A critical variable influencing outcomes of training is the feedback the learner receives during or following task performance. The source of feedback may be either intrinsic (arising from the individual's own sensory systems) or extrinsic (arising from an external source). The tactile information accompanying the use of a restraint is not naturally available to the performer and is classified as augmented intrinsic feedback (an external device used to increase an individual's awareness of sensory events accompanying performance).17 For reaching, it is possible that other forms of feedback may be equally or more effective than that provided by the tactile intrinsic feedback of the restraint. Augmented extrinsic feedback in the form of an auditory cue to maintain the back against the chair back during training18 is 1 alternative to the restraint. The purpose of this feasibility study was to perform a clinical, pilot investigation of the influence of auditory feedback in response to pressure (Sensor group) versus tactile feedback arising from an external device used to stabilize the trunk (Stabilizer group) for training reaching in individuals with stroke. Once the feedback was no longer in place, it was expected that the participants would exhibit having learned to reach with greater use of their arm and less use of the trunk, minimizing dependency on the feedback. On the basis of prior findings,3,15 we hypothesized that the auditory feedback about pressure would be associated with greater improvement in functional reaching compared to the tactile feedback of the restraint.
Participants were individuals with chronic unilateral stroke (≥6 months) who had moderately severe to severe impairment of hand function. All gave informed consent to participate in the study, which was approved by the institutional review boards at the University of the Sciences in Philadelphia (USP) and Magee Rehabilitation Hospital. Letters describing the investigation were sent to individuals with stroke who had been discharged from all rehabilitative services at Magee Rehabilitation Hospital at least 6 months earlier. Flyers were posted at all Magee Rehabilitation Hospital settings describing the investigation and calling for participants. Attending physiatrists and therapists were given an in-service describing the benefit of additional therapy to their patients at a time when these individuals were no longer eligible for services.
Individuals who chose to be evaluated for the investigation contacted the investigator via phone, and initial screenings were set up to determine whether exclusion criteria were present. Exclusion criteria were receptive aphasia, apraxia, or other cognitive deficits as determined by the investigator during interview question and answer, as well as command following during screening sessions. Additional exclusions were (1) Motor Assessment Scale (a classification of arm and hand movements using the involved upper extremity)19 score of fewer than 4 for upper arm and less than 1 for hand (these cutoff levels allow inclusion of participants who have scapular stability/mobility and some grasping activity), (2) sensory/perceptual or orthopedic problems that would limit ability to reach across a tabletop without sliding the arm on the surface, and (3) any cardiac or pulmonary conditions that would limit participation in exercise (physician clearance was required).
Participants were subsequently further screened for inclusion criteria, including moderate or severe impairment of arm movement, defined as scores of 20 to 44 on the upper-arm subsection of the FMA (a series of functional multitask arm/hand movements), out of a possible score of 66.16 Individuals scoring less than this range have previously been determined to be unable to participate in the training of the type used in this investigation,3 and individuals scoring greater than 50 are considered to be only mildly impaired.10 Before further participation in the study, written medical clearance was obtained from the participant's physician.
Of the potential participants screened, 16 individuals met the criteria and were randomly assigned to either the auditory Sensor feedback group or the Stabilizer feedback group (Table 1). All participants who began the training completed it. The investigator, a licensed physical therapist, performed all screening and testing but was not involved in the randomization of the participants into each of the training groups. Research assistants who were not familiar with the research prior to the investigation randomly assigned participants as they enrolled in the study. Participants were not blinded to their group assignment but were unaware of the study hypothesis or the primary outcome measures.
Five physical therapy students in the final year of their Master of Physical Therapy program and 1 recently graduated physical therapy student were engaged as voluntary “trainers” in the research project. All trainers successfully completed the NIH protocol on responsible conduct of human research. The investigator was not blinded to participant group assignment because of the need for the trainers to be supervised, which required the investigator to be present at various points of the training.
Pre- and Posttest Outcome Measures
Reaching and upper limb function were assessed with trunk motion unrestrained. A battery of standardized outcome measures was administered 5 or fewer days prior to the start of training (pretest) and 2 or fewer days following the completion of training (posttest). Outcome measures were selected with the intent to ensure that all domains of the International Classification of Functioning, Disability and Health model were represented,20 wherein body structure and function measures assess motor impairment, activity is measured by daily task performance, and participation is measured by life situations.
The Reaching Performance Scale (RPS) is a measure of how the arm moves relative to the trunk, as well as the extent of trunk movement used in reaching and lifting an object set on the table directly in front (near) and then 30 cm from the edge (far).21 The RPS is classified as a body function/structure measure22 that assesses changes in impairment of the arm and trunk, and measures the timing, smoothness, and directness of arm movement.21 The FMA is also a body function/structure measure that assesses the ability to perform multiple, active, resisted, and coordinated movements of the involved upper extremity,12 active range of motion (AROM; via goniometry) of the elbow and shoulder,23 and grip strength (via dynamometry).24
The Wolf Motor Function Test score represents the total time to complete a number of tasks with 1 or both upper extremities and is performed while sitting at a table.25 The WMFT, while considered both an impairment measure and an activity measure, was used in this study as the activity measure because it assessed functional performance of task-based activities. Lastly, the Motor Activity Log (MAL) was the participation measure, as it represents the participant's perception of function on a number of everyday activities involving the upper extremities.19
Participants were trained 2 to 3 days/weeks in 40- to 45-minute sessions for a total of 12 sessions. Participants assigned to the Stabilizer group were seated with trunk motion restrained by a chest harness, which was attached to the chair's back and had 2 padded shoulder straps that came across the chest (Figure 1). Participants in the Sensor group were instructed to focus on keeping their backs against the sensor that was adhered to the cushion of the chair back. These sensors were connected to an auditory feedback signal that provided extrinsic feedback when forward trunk movement removed the pressure of the trunk from the sensor (Figure 2).
Intermittent feedback for both groups was provided according to a faded feedback protocol in which the feedback was provided for 80% of the trials in training sessions 1 to 4, 60% of the trials in training sessions 5 to 8, and 40% of the trials in training sessions 8 to 12. Feedback was structured on the basis of the guidance hypothesis,26 wherein the beneficial effects of augmented feedback for reducing error can ultimately impede learning if the learner comes to depend on the feedback. This protocol of fading the frequency of feedback was developed on the basis of previous investigations that demonstrated that progressive fading of feedback over every one-third trials (over the course of training) was effective in promoting motor learning.27,28
Participants in both groups performed 150 to 200 movements, using the paretic arm in each training session. Some participants were able to achieve only 150 movements at the start of training but eventually worked up to more repetitions. Other participants were able to achieve close to 200 movements at the start of training and therefore required more complex tasks to work at the highest level of their ability during the 1-hour training session.
Participants were instructed to move at their preferred speed. Preferred speed was selected because, during pretraining trials using the pressure sensor device, errors rates were excessive when participants were required to adhere to an imposed speed criterion. Furthermore, while factors such as intensity of practice, motivation, repetition, and variability have been found to influence poststroke motor rehabilitation in TRT investigations,29 speed has not been shown to be a critical factor. However, once training progressed and participants began to find the tasks easier, they were encouraged to increase performance speed. If a participant completed the required training tasks in less than 45 minutes, then subsequent sessions employed more complex reaching tasks.
Participants were seated in a standard chair with no armrests. The chair was placed at a table with a large workspace and positioned so that there was a clearance of approximately 10 cm (4 inches) between the participant's body and the edge of the table. Objects were placed in different locations on the tabletop so that reaches of various directions and amplitudes were required to contact or grasp them. Participants were able to move the arm anywhere objects were placed, as all objects were placed within the arm's length.
Training activities progressed from reaching to contact objects, then reaching and grasping, and eventually transporting the objects. Common objects were used (eg, cups, mugs, writing, and eating utensils) and the objects varied in size, shape, and weight. Training also included unimanual activities such as sliding the arm across the tabletop with objects in the hand or tossing balls into boxes placed throughout the workspace. Bimanual activities included using 1 hand to pour water or open a container, while the other hand stabilized the object, moving objects in/out of containers, bimanual cutting activities, and folding activities. Training activities were adapted from literature related to reaching to grasp theories.30
A 2 × 2 (Group × Time) analysis of variance with repeated measures on the factor Time was used to determine whether there were significant between-groups differences in the pretest/posttest change in outcome measures.31 Significant differences were further tested using post hoc comparisons between groups, according to Tukey's HSD (honestly significant difference) Test.32 Significance for all tests was set at P < 0.05. Levene's Test of Equality of Variance was used to confirm whether the data were normally distributed.
Age and pretest scores on outcome measures were not significantly different between groups. Mean age for the Sensor group was 62.9 ± 6.5 years and for the Stabilizer group was 63 ± 9.2 years. There were no differences in pretest scores between groups on the RPS-near (P = 0.53), RPS-far (P = 0.88), FMA (P = 0.24), WMFT (P = 0.56), shoulder flexion (P = 0.71), or elbow extension (P = 0.24).
For the RPS-near target, there was a significant main effect of Time (P < 0.01). There was also a significant Time × Group interaction for this measure (P < 0.04; see Figure 3). Post hoc analysis indicated that these differences were explained by the significantly greater improvement for the Sensor group compared to that for the Stabilizer group. For the RPS-far target, there was a significant main effect (P < 0.01) of Time, with no significant interaction effect.
For the FMA scores and the WMFT scores (P < 0.04), there was a significant main effect of Time (P < 0.02). For shoulder flexion AROM, there was a trend toward a main effect for improvement (P = 0.059). Measures of elbow extension (P = 0.1), grip strength (P = 0.69), and MAL (amount of use scale; P = 0.78) yielded no significant changes. All data are shown in Table 2.
Results indicated that faded sensory feedback in the form of both extrinsic auditory feedback (Sensor group) and intrinsic tactile feedback (Stabilizer group) yielded changes in impairment and activity measures. Improvements were observed in reaching (both RPS-near and -far), shoulder AROM, FMA, and WMFT for these individuals with moderate or severe arm impairments due to chronic stroke. However, reaching in the immediate workspace (RPS-near) improved more with auditory feedback.
Although previous work has demonstrated positive effects of using trunk restraint to promote improved arm use,9,15 there has been minimal information on the value of using auditory feedback from a pressure sensor device.17 The improvements made in the Sensor group further support the use of an external device to increase awareness of sensory events that accompany movements while learning a motor task. One systematic review evaluating the use of augmented feedback during rehabilitation (primarily in individuals with stroke) concluded that there was no firm evidence supporting its effectiveness for improving motor function of the upper extremity.33 The main reason for this finding was that the longer-term effects of auditory feedback were often not investigated in the studies available for review. In a more recent review of the effect of extrinsic feedback on motor learning in the upper limb poststroke, auditory feedback was found to improve movement quality immediately following training.18
The RPS measures the timing, smoothness, and directness of reaching movements.21 The finding that reaching performance within the participant's immediate workspace significantly improved without a trunk restraint (an approach that has dominated the literature in this area) supports the value of utilizing some form of auditory feedback for reaching to grasp activities.9,15 It is possible that reliance on an external trunk stabilizer to place pressure on the anterior shoulder during training may limit the ultimate goal of this training, which is to reach out and away from the body with the impaired arm, thereby allowing for more use and function during everyday tasks. With the use of auditory feedback from a pressure sensor, a physical restraint does not prevent movement from occurring; rather, the individual makes a cognitive decision to control the trunk in response to an auditory feedback signal.
In explaining movement recovery, it is important to know whether gains in movement (both performance and outcome) are due to recovery or compensation.22 Recovery is the reappearance of premorbid movement patterns while compensation is the appearance of alternative movement patterns.6,22 The population studied in the present investigation, classified as having moderate or severe impairment, has typically gained functional improvement by compensatory methods, such as greater trunk displacement/rotation, scapular elevation, shoulder abduction, and internal rotation.34,35 Physically restricting trunk movements to gain upper limb movement has been effective in improving motor patterns and limiting compensatory strategies in less-impaired individuals.3,15 The outcomes of the present study indicate that individuals poststroke with moderate or severe arm impairments have the potential to exhibit recovery with training that incorporates auditory feedback. The similar positive results for the majority of outcome measures in this investigation (for both extrinsic auditory feedback and intrinsic tactile feedback) suggest that the common component of this training, TRT with faded sensory feedback, is the primary factor underlying improvements. These findings and our prior findings3,15 indicate that TRT is associated with improvements regardless of whether training was carried out with or without trunk-restraint. Recent reviews of the evidence for stroke rehabilitation and recovery demonstrate strong support for TRT focusing on repetitive practice of meaningful tasks of progressively increasing difficulty.36 When the trunk is free to move during reaching, individuals poststroke with more severe impairments of arm function use trunk motion to compensate for inadequate control of the upper limb.3,9,13,34 This suggests that if the individual is not trained to control the trunk, those with moderate or severe impairments of arm function will make more compensatory movements during reaching activity. Current clinical opinion dictates that compensatory strategies should not be encouraged as these movements are associated with problems such as pain, discomfort, and joint contractures.4,37
The fact that the investigator who performed the data collection and analysis was not blinded to the participant treatment group is a limitation of this investigation. In addition, the sample size of this study was small, which may have masked significant group effects for measures such as the WMFT and the MAL, that have great amounts of variability. Prior studies have shown training-related changes in MAL scores in individuals with more severe impairment than participants in the current study.39,39 Furthermore, a qualitative component (beyond the MAL) asking participants about their experiences and perceptions about the program and its results may have captured significant changes in participation. It is the opinion of the author that the Motor Assessment Scale has floor/ceiling effect, limitations and thus was used in this investigation only as an exclusion criteria.
Studies with additional analyses and long-term follow-up are needed. To answer questions about the value of auditory pressure feedback, information about long-term retention of improvements in arm function, movement quality, and decreases in compensatory trunk movement are necessary. Thus, more sophisticated movement analysis and extended follow-up of these participants are important. Measures to identify neuroplastic changes related to this type of training, such as a functional magnetic resonance image, would further support the value of this approach. Future investigations should address these issues.
In individuals with moderate or severe arm impairment poststroke, TRT that incorporated auditory sensory feedback about trunk movement was associated with improved reaching and functional use of the hemiparetic arm. While most results were comparable to those obtained with training that incorporated trunk stabilization, improvements in reaching in the immediate workspace were greater with the auditory feedback. Extrinsic feedback training should be considered for individuals poststroke. Based on these pilot findings, future randomized controlled investigations would be indicated to further evaluate training protocols that utilize fading of feedback for the stroke population.
We thank the following USP graduate students who assisted in the training of the participants: Kate Bartnik, Ravi Buddharaju, Pat Hennessy, Julie Kametz, Stacy Prokopchuk, and Kim Wallace.
1. Ward NS, Cohen LG. Mechanisms underlying recovery of motor function after stroke. Arch Neurol. 2004; 6:1844–1848.
2. Page SJ, Sisto S, Levine P, Johnston MV, Hughes M. Efficacy of modified constraint induced movement therapy in chronic stroke: a single-blinded randomized control trial. Arch Phys Med Rehabil. 2004; 85:14–18.
3. Thielman GT, Dean CM, Gentile AM. Rehabilitation
after stroke: task-related training vs. progressive resistive exercise. Arch Phys Med Rehabil. 2004; 85:1613–1618.
4. Ada L, Canning CG, Carr JH, Kilbreath SL, Shepherd RB. Task-specific training of reaching
and manipulation. In: Bennet KM, Castiello U, eds. Insights Into the Reach to Grasp Movement. Amsterdam, the Netherlands: Elsevier; 1994:239–264.
5. Gentile AM. Skill acquisition: action, movement, and neuromotor processes. In: Carr JH, Shepherd RB, eds. Movement Sciences: Foundation for Physical Therapy in Rehabilitation
. 2nd ed. Gaithesburg, MD: sp; 2000;111–187.
6. Dean CM, Richards CL, Malouin F. Task-related circuit training improves performance of locomotor tasks in chronic stroke: a randomized, controlled pilot trial. Arch Phys Med Rehabil. 2000; 81:409–417.
7. Dean CM, Shepherd RB. Task-related training improves performance of seated reaching
tasks after stroke. Stroke. 1997; 28:722–728.
8. Wu C, Trombly CA, Lin K, Tickle-Degnen L. A kinematic study of contextual effects on reaching
performance in persons with and without stroke: influences of object availability. Arch Phys Med Rehabil. 2000; 81:95–101.
9. Michaelson SM, Levin MF. Short-term effects of practice with trunk restraint on reaching
movements in patients with chronic stroke. Stroke. 2004; 35:1914–1919.
10. Michaelson SM, Dannenbaum R, Levin MF. Task-specific training with trunk restraint on arm recovery in stroke: randomized control trial. Stroke. 2006; 37:186–192.
11. Richards L, Pohl P. Therapeutic interventions to improve upper extremity recovery and function. Clin Geriatr Med. 1999; 15:819–832.
12. Lincoln N, Leadbitter D. Assessment of motor function in stroke patients. Physiotherapy. 1979; 65:48–51.
13. Cirstea MC, Levin MF. Compensatory strategies for reaching
in stroke. Brain. 2000; 123:940–953.
14. Thielman GT, Gentile AM. Rehabilitation
post stroke: follow up assessment. Soc Neurosci Abstract. 2002; Orlando FL.
15. Thielman GT, Kaminski TR, Gentile AM. Training with trunk restraint during reaching
for individuals post stroke. Neurorehabil Neural Repair. 2008; 22:697–705.
16. Gladstone DJ, Danells CJ, Black SE. The Fugl-Meyer assessment of motor recovery after stroke: a critical review of its measurement properties. Neurorehabil Neural Repair. 2002; 16(3):232–240.
17. Magill AR. ed. Motor Learning and Control: Concepts and Applications. 8th ed. New York, NY: McGraw-Hill; 2006.
18. Subramanian SK, Massie CL, Malcolm MP, Levin MF. Does provision of extrinsic feedback result in improved motor learning in the upper limb poststroke? A systematic review of the evidence. Neurorehabil Neural Repair. 2010; 24:113–124.
19. Carr JH, Shepherd RB, Nordholm L, Lynne D. Investigation of a new motor assessment scale for stroke patients. Phys Ther. 1985; 65:175–180.
20. World Health Organization. International Classification of Function in Disability and Health: ICF. Geneva, Switzerland: World Health Organization; 2001.
21. Michaelson SM, Jacobs S, Roby-Brami A, Levin MF. Compensation for distal impairments of grasping in adults with hemiparesis. Exp Brain Res. 2004; 57:162–173.
22. Levin MF, Kleim JA, Wolf SL. What do motor “recovery” and “compensation” mean in patients following stroke? Neurorehabil Neural Repair. 2009; 23(4):313–319.
23. Norkin C, White D. Measurement of Joint Motion: A Guide to Goniometry. Philadelphia, PA: FA Davis Company; 1995.
24. Harris JE, Eng JJ. Paretic upper-limb strength best explains arm activity in people with stroke. Phys Ther. 2007; 87(1):88–97.
25. Wolf SL, Catlin PA, Ellis M, Archer AL, Morgan B, Piacentino A. Assessing wolf motor function test as outcome measure for research in patients after stroke. Stroke. 2001; 32:1635–1639.
26. Schmidt R, Lee DL. Motor Control and Learning: A Behavioral Emphasis. Champaign, IL: Human Kinetics; 2005.
27. Cirstea MC, Levin MF. Improvement of arm movement patterns and endpoint control depends on type of feedback during practice in stroke survivors. Neurorehabil Neural Repair. 2007; 21:398–411.
28. Cirstea CM, Ptito A, Levin MF. Feedback and cognition in arm motor skill reacquisition after stroke. Stroke. 2006; 37:1237–1242.
29. Kleim JA, Jones TA. Principles of experience-dependent neural plasticity: implications for rehabilitation
after brain damage. J Speech Lang Hear Res. 2008; 51:S225–S239.
30. Ada L, Canning C, Carr J, et al. Task Specific training of reaching
and manipulation. In: Bennett KMB, Castiello U, eds. Insights Into the Reach to Grasp Movement. Amsterdam, the Netherlands: North Holland; 1994:239–265.
31. Judd CM, McClelland GH. Data Analysis: A Model-Comparison Approach. San Diego, CA: Harcourt Brace College Publishers; 1989.
32. Portney LG, Watkins MP. Foundations of Clinical Research: Applications to Practice. Upper Saddle River, NJ: Prentice Hall; 2008.
33. van Dijk H, Jannink MJA, Hermens HJ. Effect of augmented feedback of motor function of the affected upper extremity in rehabilitation
patients: a systematic review of randomized controlled trials. J Med Rehabil. 2005; 37:202–211.
34. Levin MF, Michaelson S, Cirstea CM, Roby-Brami A. Use of the trunk for reaching
targets placed within and beyond the reach in adult hemiparesis. Exp Brain Res. 2002; 143:171–180.
35. Roby-Brami A, Jacobs S, Bennis N, Levin MF. Hand orientation for grasping and arm joint rotation patterns in healthy subjects and hemiparetic stroke patients. Brain Res. 2003; 969:217–229.
36. Teasell R, Bayona N, Salter K, Hellings C, Bitensky J. Progress in clinical neurosciences: stroke recovery and rehabilitation
. Can J Neurol Sci. 2006; 33(4):357–364.
37. Levin MF. Should stereotypic movement synergies seen in hemiparetic patients be considered adaptive? Behav Brain Sci. 1997; 19:79–80.
38. Page SJ, Harnish SM, Larmy M, Eliassen JC, Szaflarski JP. Affected Arm Use and Cortical Change in Stroke Patient's exhibiting Minimal Hand Movement. Neurorehabil Neural Repair. 2010; June (5): 495–502.
39. Bonifer NM, Anderson KM, Arciniegas DB. Constraint-induced movement therapy after stroke:efficacy for patients with minimal upper extremity motor ability. Arch Phys Med Rehabil. 2005. Sep;(86) 9: 1867–1873.