Healthy adults are remarkably adept at manipulating a diversity of objects, adjusting grip forces to various object properties with limited conscious awareness of the properties themselves or the actual adjustments that are made in response to those properties. For example, when an object is held unsupported in space by the anatomical hand (AH), grip forces are subconsciously adjusted according to the object’s weight1–4 and surface friction.5,6 However, due to the lack of afferent sensory modalities within the prehensor itself, this constant calibration may be lacking for upper-limb prosthesis users.
Tactile sensory feedback from the skin of the grasping fingertips of the AH plays a crucial role in the automatic control of prehensile finger forces, providing real-time sensory feedback about changes in inertial loads and the stability of the interface between the skin surface and the object as the object is held or transported.1–4,7,8 Tactile skin receptors signal important information about movement related events, including that the object has been contacted or that it has started to move or to slip from the grasp. The importance of tactile sensory information in fine manipulative movements is supported by the problems commonly experienced by persons with impaired manual sensibility who frequently drop objects, easily crush fragile objects, and have problems performing simple activities of daily living (ADL) such as buttoning a shirt when dressing.9–11 Such observations led Moberg (1975) to claim that reconstructive surgery of the hand should aim to preserve all possible cutaneous innervation because the smallest grip with sensibility is preferable to the best hand prosthesis.
Research on grip force control with the AH suggests that cutaneous sensory cues are important for modulating force when manipulating an object. Upper-limb prosthesis users are not able to take advantage of sensory cures to modulate force when manipulating an object, at least at the level of the hand. If, as suggested by contemporary researchers, effective grip force regulation during object manipulation results from an internal model that is updated by peripheral sensory information, then the person with amputation has two major limitations that will reciprocally influence each other. First, during the early stages of skill acquisition with their new prosthesis, the person with amputation will not have an accurate internal model of interactions among limb dynamics, object properties, and task constraints to specify appropriate neural commands to the body parts that control the prosthesis and its prehensor. Second, the person with amputation will lack the cutaneous sensibility at the prehensor, which is needed to pick up real-time information about changes in the inertial characteristics of the object and potential changes at the interface between the object surface and the grasping surface of the prehensor as the object is held or moved around. Ultimately, the lack of cutaneous sensibility could significantly retard or even prevent the development of effective internal models for grip force regulation during object manipulation.
Although accurate statistics are illusive, by some conservative estimates, there are over 100,000 people with upper-limb amputations in the United States.12,13
However, the Amputee Coalition14 in conjunction with the National Limb Loss Resource Center15 compiled statistics that suggest the number of individuals with upper-limb amputations may exceed prior assumptions. Most upper-limb amputations are below elbow, and a greater percentage of people with amputations use body-powered artificial limbs, as opposed to externally powered prostheses such as myoelectrically controlled devices.16
Two types of prehensors, which function in an inverse manner relative to each other, are typically used with the body-powered prosthesis. The voluntary opening (VO) device is naturally closed and requires an increase of cable tension to open. The voluntary closing (VC) device is naturally open and requires an increase of cable tension to close. A recent study investigated the functionality and preference of users of the VO and VC prehensors in a validated test of ADL.17 The results of the study indicated that the VC device allowed task completion to be 1.3 seconds faster on average than the VO, although no clear differences in functionality were observed. Some participants preferred to use the VC, whereas others preferred the VO, depending on the type of task. Some participants expressed concern that the VO lacked sufficient grip force in some activities. This limitation is due to the fact the maximum grip force in the VO prehensor is determined by the strength of the rubber bands that are used to close the prehensor. Interestingly, most participants commented that it was more difficult to perceive the force they were applying to an object using the VO prehensor compared with the VC. One possible explanation for this finding is that participants need to decrease the effort they use when applying force to an object with the VO prehensor, whereas they do just the opposite with the VC prehensor. The different relation between muscle tension and grip force has been discussed previously relative to the notion of compatibility. Based upon a large literature on stimulus-response (S-R)18 and response-response (R-R)19 compatibility in the psychological and motor control literature, as well as observations from others in the field of prosthetics,20–23 it has been suggested that the relation between muscular tension and grip force may be less intuitive in the VO prehensor than the VC prehensor, at least for new prosthesis users.24 Because of the lack of direct sensory feedback from the prehensors (there are no force receptors on the prehensors), grip force is not directly fed back to the user. Rather, the user perceives grip force indirectly from the muscle tension used to apply the force. Thus it would seem that the relationship between the muscle tension used and the resulting changes in prehensor grip force should be intuitive to maximize performance.
The use of a bicycle brake analogy helps describe this type of intuitive relationship or compatibility. To brake a bicycle, the bicyclist applies grip force, causing an increase in the tension on the brake cable and the brake pads to close, slowing down or stopping the bicycle. This operation represents a positive correlation, and possibly an intuitive or logical compatibility between grip force and brake force. The VC prehensor shares this type of compatibility because generating muscle force by body motions increases tension on the cable that connects the shoulder harness system and the prehensor, causing prehensor grip force to increase. The correlation between muscle tension and grip force is positive in the VC prehensor. In contrast, the correlation between muscle tension and prehensor grip force in the VO prehensor is negative because, to increase prehensor grip force, the user must reduce cable tension by body motions. Consequently, the VO and VC prehensors might be suited to different types of tasks.24
To examine this idea, we previously conducted a case study with a person having congenital, quadruple amputations who uses both upper- and lower-limb prostheses on all four limbs.25 This individual was an experienced prostheses user who only used body-powered upper-limb VO prehensors for both right and left limbs and who had no previous experience with a VC prehensor. The unique manipulation in this study was that we asked our participant to perform a grip force control task, similar to the one used in the current study, with his own VO prosthesis and also with a VC prosthesis that he had never used before. On some measures of grip force control, the participant actually performed better with the unfamiliar VC prosthesis.
To further examine this hypothesis, in the present experiment, we asked a group of participants without amputation to perform a similar grip force task using an upper-limb body-powered prosthetic simulator26 that mimics the functionality of a typical body-powered prosthesis used by people with amputations. Participants performed the grip force control task with the VO and VC prehensors fixed to the prosthetic simulator with and without augmented visual feedback of the target forces they were asked to produce. We wished to determine whether the predicted effects of muscle tension–grip force compatibility would be present with and without augmented visual feedback. The augmented visual feedback was initially provided to help the participant produce the required grip force, and then it was removed to see if the participant could maintain the level of grip force (Figures 1, 2).
Based on the prior research with the experienced prosthesis user, we predicted that participants would produce the target force more accurately and with less variability when they used the VC prehensor compared with the VO prehensor. In addition, we expected that, due to experience, performance with the AH would be superior to performance with the VO and VC prehensors.
MATERIALS AND METHODS
Ten able-bodied individuals (average age, 22 ± 4 years; 5 men and 5 women) participated in the study. Eight participants were right hand dominant, and two participants were left hand dominant. All participants gave informed written consent for this study that was approved by the institutional review board at San Francisco State University.
Grip force was measured using Biopac MP30 Hardware for Biopac Student Laboratory Pro software along with the Biopac Systems hand dynamo set at a sampling rate of 500 samples per second. The hand dynamo was placed on a wooden platform that allowed the dynamo to stand erect in a locality that was comfortable for the participant to grasp. The volunteer was situated so that he/she could grasp the dynamo with the terminal device of the prosthetic simulator, which was projected forward in the sagittal plane (Figure 3). A force output tracing was displayed on a 16-inch monitor that was placed approximately 1 m away from where the participant sat upon a stationary chair. The body-powered prosthetic simulator, developed in-house, could be equipped with either a Sierra 2-Load VO hook (Hosmer Dorrance Corp; Campbell, CA, USA) or an APRL VC hook (Hosmer Dorrance Corp) and utilized a cable-harnessing system identical to an actual prosthesis. Two simulators, one for left and one for right hand dominant participants, were utilized.
Before the participant sat down to perform the experiment, he/she watched an informational video created by our research team. The video consisted of a demonstration and commentary describing how terminal device manipulation was accomplished through manipulation of cable tension. The informational video demonstrated in an explicit manner the difference between the two types of terminal devices and what actions would open or close each device. The participant was then asked if he/she understood how to increase, decrease, and control grip force with both devices. All participants confirmed that the instructional video provided sufficient explanation for prehensor control. The experimental protocol was then read to the participants, and they were asked if they were ready to begin.
After viewing the demonstration video and confirming readiness, the volunteer donned the prosthetic simulator on their dominant side with help from the experimenter. The participant was situated in a chair so that he/she could comfortably grasp the hand dynamo. Three practice trials were given before the experiment began. Each participant received practice with the VO, AH, and VC in that order. In the practice trials, the participants produced grip forces of 1 N, 5 N, and 12 N all with the use of vision. The participants were then informed that they were about to begin the 54 trial protocol and asked if they were ready to begin. After confirmation, the experimental protocol began.
Each trial lasted for a total of 15 seconds. In the experimental trials, the participant was given 5 seconds from the start of each trial to reach the target grip force with the aid of augmented visual feedback by manipulating grip force on the dynamometer. He/she was then asked to sustain the target grip force for 10 seconds on each trial. In the no augmented visual feedback condition, visual feedback was manipulated by covering the monitor screen during the last 10 seconds of the trial. In the augmented visual feedback condition, the monitor was uncovered during the last 10 seconds of the trial. We believed that this type of grip force task, similar to the one used previously,25 would allow us to adequately examine the hypothesis under different visual feedback conditions. This type of task has commonly been used to study the control of grip forces during grasping, to identify the brain areas involved in the regulation of grip forces, and to measure deficits in grip force control in patients with various central nervous system pathologies.26
The experiment consisted of a 3 × 2 × 3 (hand condition × augmented visual feedback condition × target force) design. Hand condition consisted of performing the target grip force task with the VO prehensor, the VC prehensor, and the AH. Augmented visual feedback condition consisted of performing the target grip force task with or without augmented visual feedback for the last 10 seconds of each trial. The target force conditions were either 0.49 N, 4 N, or 10.5 N. Each unique hand condition, augmented visual feedback condition, and target force condition was repeated three times in a row for a total of 54 trials. Intertrial interval was approximately 15 seconds. The participants fell into one of two groups. The first group, n = 5, performed all conditions with the VC split hook, then the AH, then the VO. The second group, n = 5, performed all conditions with the VO, then the AH, then the VC. Unfortunately, the data for three subjects were compromised because of a hardware malfunction. The second and tertiary conditions were randomized for each hand condition. For example, the presence or absence of augmented visual feedback as well as force variables was randomized. After all 18 trials were completed for each hand condition, the participant was given an opportunity to rest before continuing testing with the next effector protocol. The fatigue effect was reduced by interceding the two simulator conditions with the AH condition.
Three measures of grip force accuracy and variability were assessed for each condition. Absolute constant error is a measure of accuracy and is defined as the absolute value of the difference between the target force and the force output. Variable error is a measure of variability and is defined as the standard deviation in the difference between target force and force output. Biasing index (BI) is a measure that quantifies the degree to which the participant overshot or undershot the target and can provide more details of the types of grip force errors the participant can make. BI is calculated as:
BI has a range of −1 to 1, and values greater than 0 indicate a bias toward overshooting a target force, and values less than 0 indicate a bias toward undershooting a target force. For example, if the participant consistently exerted either too little or too much grip force relative to the target force over a set of trials, the biasing index would be significantly less or greater than zero, respectively.
A one-way analysis of variance (ANOVA) was performed for each force × vision condition to assess performance differences for each condition. The primary question we wanted to answer was how each prosthetic prehensor (VO vs. VC) performed relative to each other as well as to the AH in producing a target grip force both with and without augmented visual feedback.
ABSOLUTE CONSTANT ERROR
For absolute constant error, larger values indicate less accuracy and lower values represent greater accuracy. Figure 4 represents mean ± standard error measure for the absolute constant error for each force × augmented visual feedback condition. One observation gleaned from this figure is that errors in the no-augmented visual feedback conditions (bottom panel) are generally larger than under augmented visual feedback conditions (top panel). Please note that the vertical scale (in Newtons) is much smaller for the augmented visual feedback condition graphs than for the no-augmented visual feedback condition graphs. This result clearly points to the importance of augmented visual feedback for this type of task requiring the matching of actual grip force to the target grip force. The augmented visual feedback clearly provided information that facilitated error detection and correction.
In addition, closer inspection reveals that, particularly in the no-augmented visual feedback conditions, the pattern of errors for the AH and VC are more similar than for the AH and the VO. Generally, for the AH and VC conditions, errors increase as the target forces increase. However, for the VO condition, errors decrease as the target force increases.
Although the absolute constant errors were very small in the augmented visual feedback condition, a one-way within condition ANOVA revealed a significant difference in absolute constant error for the 4 N [F(2,60) = 3.73, P < 0.05] and 10.5 N [F(2,60) = 6.8, P < 0.05] augmented visual feedback conditions, as well as the 0.49 N [F(2,60) = 9.02, P < 0.001] and 10.5 N [F(2,60) = 4.7, P < 0.05] no-augmented visual feedback conditions. A Tukey’s honestly significant difference post hoc test revealed that the VO prehensor was significantly less accurate than the VO and AH for the 4 N target force with augmented visual feedback. In addition, the AH was significantly less accurate than both the VO and VC for the 10.5 N target force with augmented visual feedback.
Without augmented visual feedback, the VO prehensor was less accurate than both the AH and VC for the 0.49 N target, but the AH was significantly less accurate than the VO for the 10.5 N target force.
Higher variable error means more variability in performance and lower variable error represents more consistency in performance. Similar to absolute constant error, variable errors were generally higher in the no-augmented visual feedback compared with the augmented visual feedback conditions. In addition, the patterning of errors, particularly in the no-augmented visual feedback conditions were more similar for the AH and VC conditions compared with the AH and VO conditions.
Figure 5 represents mean ± standard error measure for the variable error for each force × augmented visual feedback condition. A one-way within condition ANOVA revealed a difference in variable error for the 0.49 N [F(2,60) = 4.6, P < 0.05] and 10.5 N [F(2,60) = 6.18, P < 0.01] target forces with augmented visual feedback as well as the 0.49 N [F(2,60) = 12.44, P < 0.001], 4 N [F(2,60) = 3.71, P < 0.05] and 10.5 N [F(2,60) = 7.12, P < 0.01] target forces without augmented visual feedback. A Tukey’s honestly significant difference post hoc test revealed that the VO was significantly more variable for the low force condition (0.49 N) with augmented visual feedback. The AH was significantly more variable than both the VC and VO condition for the high target force (10.5 N) with augmented visual feedback.
Without augmented visual feedback, the VO was significantly more variable than both the AH and VC for the 0.49 N condition. The VO was also more variable than the VC for the 4 N condition. The AH was significantly more variable than both the VC and VO for the 10.5 N target force condition.
Figure 6 represents mean ± standard error measure for biasing index. A one-way within condition ANOVA revealed a difference in biasing index for the 4 N [F(2,60) = 17.53, P< 0.001] and 10.5 N [F(2,60) = 9.24, P < 0.001] target force with augmented visual feedback and 4 N [F(2,60) = 13.19, P < 0.001] and 10.5 N [F(2,60) = 7.03, P < 0.005] target force without augmented visual feedback. A Tukey’s honestly significant difference post hoc test revealed that both the VO and VC overshot the 4 N target force with augmented visual feedback, but the AH slightly undershot the target force. The VO overshot the target more than the VC, and the VC and AH were more similar. For the 10.5 N target force with augmented visual feedback, both the AH and VC undershot the target on average, but the VO was closer to the target than both the VC and AH.
For the no-augmented visual feedback conditions, both the AH and VC undershot the target for the 4 N target force, but the VO overshot the target. For the 10.5 N target force, all three prehensors undershot the target force on average. It should be noted that the only difference between the VC and AH was for the 4 N target force with augmented visual feedback. These results indicate that the trend toward undershooting and undershooting the target is the same for the AH and VC.
The purpose of the present study was to assess the accuracy and variability of grip force control using a VO and VC prehensor and the AH. Participants were given augmented visual feedback initially to help them produce the target force and then the augmented visual feedback was either maintained or removed to assess how well the target force could be maintained. Our results indicated that when augmented visual feedback was available, there were little accuracy and variability differences in the types of prehensor used in the grip force task, although some statistically significant differences between the conditions occurred. The accuracy and variability differences between the prehensor conditions were less than 0.3 N and 0.5 N, respectively. We would interpret these results by saying that when augmented visual feedback is available, no meaningful performance differences occurred between the prehensors. This result suggests that visual feedback might be used to overcome any limitations of the prehensors. Augmented visual feedback allows the participant to compare their exerted grip force with the target grip force on the computer monitor so that any discrepancies between them can be corrected. This type of knowledge of results has been shown to be a powerful type of feedback for skill acquisition.27 So, when augmented visual feedback was available, there were no meaningful performance differences between the two prehensors. However, it is important to note that when objects are lifted, held, and manipulated during ADL, this type of feedback is not available. A person has to rely instead on intrinsic feedback provided by the proprioceptive system to maintain and adapt grip forces applied to an object.
When augmented visual feedback was removed, there were several indications that performance accuracy and variability were affected by the type of prehensor and the level of the target force. In general, and as expected, errors were greater when augmented visual feedback was unavailable. As just noted, participants needed to rely upon intrinsic feedback provided by the proprioceptive system and compare this feedback with a memory representation of the target force. With inexperienced users, it was likely that the memory of the target force based upon proprioception was relatively weak. Thus, after augmented visual feedback was removed after 5 seconds at the beginning of the trial, the participant must have tried to compare current proprioceptive feedback with a perceptual memory of the target force. Because the target force memory was likely relatively weak, and with no augmented visual feedback to rely upon, performance suffered. In general, accuracy and variability was similar for the VC and AH conditions. At the 0.49 N and 4 N target force conditions, the VC and AH conditions showed more accurate and less variable performance than the VO condition. However, at the 10.5 N target condition, the VO condition was more accurate and less variable than the VC and AH conditions. How might this result be explained?
One well-known relationship in the area of human motor control is the force-force variability relationship.28 In general, it has been found that the variability of producing muscular force increases as target force increases, whereas the exact nature of these increases (i.e., linear, exponential) has been debated.29,30 Can the force-force variability relationship provide an explanation of some of the results of the present study? If so, we would expect that as the target force increased, the variability of the grip force production using the VO, VC prehensors, and AH should have also increased. In general, our results are fairly consistent with the force-force variability relationship. However, it is important to make a distinction between the prehensor grip force and the muscular tension used to produce that force within the context of the present study. As discussed previously, with the VC and AH, an increase in muscular tension produces an increase in prehensor (or finger) grip force. Therefore, according to the force-force relationship, variability of grip forces should increase as the target grip force increases. As seen in Figure 5 (bottom panel), this result was observed, particularly in the absence of augmented visual feedback. However, in the VO condition, the target force-grip force variability relationship was effectively reversed. Figure 5 (bottom panel) shows that for the VO condition, the largest variable error was produced in the 0.49 N target force condition. One explanation is that, although 0.49 N was the smallest target force, greater muscular tension was necessary when using the VO prehensor. So, what contributed to performance variability was not the target force per se but the amount of muscle activity used to control the prehensor. This result could be further evaluated in future studies that examine the relationship between grip force control (and performance on various ADL tasks), and the type of muscle groups used to control cable tension in body-powered prostheses.
Another interesting result was that at the larger target force condition (10.5 N), performance with the AH was actually inferior to the two prostheses conditions. This result could also be explained by the force-force variability hypothesis because the muscles that control grip force in the AH are relatively smaller and weaker than those that control grip force in the prehensors (trunk and shoulder muscles). Consequently, the muscles of the AH must work at a higher relative intensity than those used to control the prehensors and thus would be expected to generate greater variability. Alternately, the superiority of the prehensors could be because of the dynamics of the cabling system. Future investigations could examine the type of oscillations produced by the VO and VC prehensor compared with the AH. Anatomical hand grip force oscillations have been shown to increase with advancing age31 and with certain neurological disorders.7 However, there is virtually nothing known about oscillations produced by prostheses and little is known whether grip force oscillations are affected by the type of prehensor (VO vs. VC, body-powered vs. EMG powered), or muscle groups being activated, for example. Clearly, much more work is needed to understand the advantages and disadvantages of using the VO and VC prehensors relative to each other and relative to the AH.
And lastly, while we feel the type of task used in the present study provided some insight into how grip pressure is regulated by different types of prehensors, more work needs to be done on whether muscle tension–grip force compatibility influences performance on different ADL tasks. This work could lead to advancing our knowledge of the strengths and limitations of the VO and VC prehensor and assist the occupational therapist in improving prosthesis training.
1. Johansson RS, Westling G. Roles of glabrous skin receptors and sensorimotor memory in automatic control of precision grip when lifting rougher or more slippery objects. Exp Brain Res
2. Johansson RS, Westling G. Signals in tactile afferents from the fingers eliciting adaptive motor responses during precision grip. Exp Br Res
3. Johansson RS, Westling G. Afferent signals during manipulative tasks in man. In: Franzen O, Westman J, eds. Somatosensory Mechanisms
. London: Macmillan; 1991:25–48.
4. Westling G, Johansson RS. Factors influencing the force control during precision grip. Exp Brain Res
5. Cadoret G, Smith AM. Friction, not tecture, dictates grip forces used during manipulation of hand-held objects. J Neurophysiol
6. Johansson RS. Sensory control of dexterous manipulation in humans. In: Wing AM, Haggard P, Flanagan JR, eds. Hand and Brain
. San Diego: Academic Press; 1996:381–414.
7. Kinoshita H, Francis PR. A comparison of prehension force control in young and elderly individuals. Eur J Appl Physiol Occup Physiol
8. Gordon AM, Duff SV. Fingertip forces during object manipulation in children with hemiplegic cerebral palsy. I: anticipatory scaling. Dev Med Child Neurol
9. Moberg E. Criticism and study of methods for examining sensibility in the hand. Neurology
10. Moberg E. Aspects of sensation in reconstructive surgery of the upper extremity. J Bone Joint Surg
11. Moberg E. Hand surgery and the development of hand prostheses. Scand J Plast Reconstr Surg
12. Atkins DJ. Adult upper-limb prosthetic training. In: Comprehensive Management of the Upper-Limb Amputee
. New York: Springer-Verlag; 1989:39–59.
13. LeBlanc M. Patient population and other estimates of prosthetics
and orthotics in the USA. Prosthet Orthot
14. Amputee Coalition. Available at: http://www.amputee-coalition.org
16. LeBlanc M. Use of prosthetic prehensors. Prosthet Orthot Int
17. Berning K, Cohick S, Johnson R, et al. Comparison of body-powered voluntary opening
and voluntary closing prehensor
for activities of daily life. J Rehabil Res Dev
18. Alluisi E, Proctor RW, Reeve TG. Things that go together: a review of stimulus–response compatibility and related effects. In: Advances in Psychology
. Amsterdam: Elsevier Science Publishers, B.V. (North Hollance); 1990;65:3–30.
19. Heuer H. Rapid responses with the left or right hand: response-response compatibility effects due to intermanual interactions. In: Proctor RW, Reeve TG, eds. Advances in Psychology
. Amsterdam: Elsevier Science Publishers, B.V. (North Hollance); 1990;65:311–342.
20. De Visser M, Herder JL. Force-directed design of a voluntary closing
hand prosthesis. J Rehabil Res Dev
21. Murphy EF. Manipulators and upper-extremity prosthetics
. B Prosthet Res Fall
22. Plettenburg DH, Herder JL. Voluntary closing
: a promising opening in hand prosthetics
. Tech Disabil
23. Radocy B. Voluntary closing
control: a successful new design approach to an old concept. Clin Prosthet Orthot
24. Wallace SA, Anderson DI, Trujillo M, Weeks DL. Upper extremity artificial limb control as an issue related to movement and mobility in daily living. Quest
25. Trujillo M, Anderson DI, Mitchell M. Grip force
using an artificial limb in a congenital amputee. Open J Ther Rehab
26. Shumway-Cook A, Woollacott MH. Motor Control: Translating Research into Clinical Practice
. 4th Ed. Baltimore: Lippincott, Williams & Wilkins; 2012.
27. Wallace SA, Anderson DI, Anderson D, et al. Motor performance with a simulated artificial limb. Percept Mot Skills
28. Magill RA, Anderson D. Motor Learning and Control: Concepts and Applications
. 10th Ed. Columbus: McGraw Hill Education; 2013.
29. Schmidt RA, Lee TD. Motor Control and Learning: A Behavioral Emphasis
. 5th Ed. Champaign, IL: Human Kinetics Publishers, Inc.; 2011.
30. Carlton LG, Newell KM. Force variability and characteristics of force production. In: Newell KM, Corcos DM, eds. Variability and Motor Control
. Champaign, IL: Human Kinetics Publishers, Inc.; 1993.
31. Sherwood DE, Schmidt RA. The relationship between force and force variability in minimal and near-maximal static and dynamic contractions. J Mot Behav