| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Neuroscience, University of Minnesota, Minneapolis, Minnesota
Submitted 9 November 2007; accepted in final form 26 January 2008
 |
ABSTRACT |
|---|
|
|
|
INTRODUCTION |
|---|
|
Previous studies have found that when objects are rotated using a thumb-index grip, grip forces change in proportion to the amplitude of rotation about a horizontal axis (Goodwin et al. 1998
; Johansson et al. 1999) and angular speed of wrist motion (Werremeyer and Cole 1997) and are linearly related to the load force at each digit. For a thumb-index grip combined with wrist flexion, Werremeyer and Cole (1997) also found that two finger muscles, first dorsal interosseus (FDI) and flexor digitorum superficialis (FDS), are recruited along with a wrist flexor, flexor carpi radialis (FCR). Thus recruitment of both intrinsic and extrinsic finger flexors may contribute to the increase in grip force observed during wrist flexion movements. Therefore even in the simplest case of a two-digit grasp, the patterns of force and muscle activity may vary depending on the specific task requirements.
If rotation of a hand-held object is performed using a three-digit tripod grip, the number of possible solutions (in terms of patterns of forces and muscle activity) increases substantially. A static tripod grasp is characterized by opposing horizontal forces that intersect at a force focus within the plane of the contact points (Baud-Bovy and Soechting 2001; Yoshikawa and Nagai 1991), and the sum of the moments with respect to the pitch, roll, and yaw axes must equal zero to prevent rotation. Therefore to produce a rotation from a static hold, characteristics of the contact forces such as direction and amplitude must be modulated at one or more contacts. Changes in static or dynamic moments of force can also be produced by altering the center of pressure locations at each contact point, resulting in a change in the moment arm through which the applied forces act on the object. Therefore force direction and amplitude as well as the center of pressure at each contact point must be coordinated to produce the desired rotation. Because rotation of a hand-held object could be achieved by using various combinations of forces, it is unclear whether specific strategies are preferred for certain rotations. Furthermore, patterns of muscle activity during rotation could reflect a contribution to the digit force and/or a general stiffening of the hand resulting in a specific force pattern due to the interaction of the object and a stiffened hand as demonstrated during object translation (Winges et al. 2007).
This study was designed to examine the extent to which unique strategies are utilized to produce rotations of a hand-held object. Specifically, we examined patterns of center of pressure, force direction, and amplitude as well as arm and hand muscle activity during hand-held object rotation using a tripod grip. We found that grip force increased monotonically for all rotations following the time course of the rotation. We also found changes in the direction of grip force and center of pressure that result primarily from the interaction of a stiff hand with a rigid object rather than from active muscular control. Activity in finger flexors was consistent with increases in grip force and thus dynamic stiffening of the hand during the rotation.
|
|
METHODS |
|---|
|
Instrumentation
Subjects grasped an instrumented manipulandum from above (described in Baud-Bovy and Soechting 2001) with the right hand, using a tripod grasp, i.e., with the thumb, ring, and index fingers on the transducers (T1, T2, and T3, respectively; Fig. 1). The contacts were constrained to the surfaces of three 17-mm diam force-torque transducers (ATI Nano 17 US-6-2) covered with No. 60 sandpaper. The transducers were arranged equidistant from the center of the manipulandum, on a circle with a radius of 42 mm such that forces normal to the contact surface of each transducer were directed toward the center of the manipulandum. The contact surface of the thumb sensor (T1) was parallel to the frontal plane of the subject. Forces and torques from the transducers were sampled at 1 kHz, and the three-dimensional orientation of the object was sampled at 60 Hz using a Polhemus Fastrak system. Force and orientation data were low-pass filtered (4th-order Butterworth, 10-Hz cut-off frequency) prior to analysis.
|
4.20 cm below the horizontal plane of the contact surfaces.
Hand muscle activity was recorded to determine the extent to which the modulation of force at each digit could be attributed to the modulation of muscle activity. [Modulation in force could also result from a mechanical interaction of a stiffened hand and the manipulandum (Winges et al. 2007).] Electromyographic (EMG) activity was recorded using small bipolar Ag/AgCl surface electrodes, with 2-mm-diam conductive surfaces placed 10 mm apart (see Fig. 2 of Winges et al. 2007). These electrodes were purchased from Discount Disposables (St. Albans, VT) and then permanently soldered to custom-made, electrically shielded wire leads, which were connected to standard laboratory amplifiers. The ground electrode was attached to the subject's contralateral, nonmoving wrist. EMG was amplified (x1,000), band-pass filtered (60–500 Hz), and then sampled at 1,000 Hz.
EMG activity was recorded from two wrist adductors: extensor carpi ulnaris (ECU) and flexor carpi ulnaris (FCU), which also produce wrist extension and flexion, respectively. Activity was also recorded from three portions of extrinsic finger muscles: a central portion of extensor digitorum (ED) and two portions of flexor digitorum superficialis (FD, FD2). The FD portion of the flexor digitorum superficialis was closer to the index finger and FD2 was closer to the ring finger. We also recorded from two intrinsic thumb muscles: abductor pollicis brevis (APB) and flexor pollicis brevis (FPB); and three intrinsic finger muscles: first dorsal interosseus (FDI) and index and ring lumbricals (LUMi, LUMr, respectively). We previously estimated cross talk between the APB and FPB electrodes to amount to <16% of the recorded signal (Klein Breteler et al. 2007), implying relatively selective recording. LUMr is beneath a relatively thick layer of tissue but is not close to other muscles; LUMi lies above adductor pollicis, but the fibers of this thumb muscle run orthogonal to the axis of the bipolar LUMi electrodes.
Experimental procedures
In response to a tone, subjects were asked to grasp the manipulandum with their right hand using a tripod grasp, lift it, perform one of six rotations, hold it, and then replace it on the table. The instructed rotation for a given trial is considered the target rotation, which was a positive or negative pitch, roll, or yaw (i.e., rotation about the transverse, sagittal, or frontal axis, respectively; Fig. 1). Other incidental (noninstructed) rotation is considered off-axis rotation. Positive and negative pitch, roll and yaw rotations are identified as forward (FWD), backward (BWD), right (RHT), left (LFT), counterclockwise (CCW), and clockwise (CW), respectively. Practice was given prior to the start of the experiment to familiarize subjects with the movements. Five subjects completed two blocks of sixty trials for each object configuration. Each block contained 10 sets of trials for six randomized object rotations. For four subjects, only force data were obtained from two blocks of 30 trials (5 sets of trials with 6 randomized object rotations), which were interspersed between blocks of object translation trials (described elsewhere, Winges et al. 2007), for each object configuration. ANOVA revealed no significant difference between the two sets of data nor did a visual inspection of the data. Therefore for purposes of analysis, we combined the data from the two experiments.
Data analysis
Orientation data were differentiated to obtain angular velocity and angular acceleration of the object during the rotation. Rotation onset and offset were determined as the points when angular velocity was 5% of the maximum for that movement. Orientation, force, and EMG data were then time normalized to 100% of the movement duration.
Local forces at each digit are defined in Fig. 1 with arrows indicating the direction of positive sign. The sum of the moments about a given axis must be nonzero to produce the respective pitch, roll, or yaw rotation. A rotation can be achieved by altering the center of pressure and/or the direction and magnitude of the force vector at one or more digits, thus multiple solutions exist to perform a single rotation. To identify common trends, force and orientation data were averaged across subjects. Results from individual subjects were then compared with the overall average. The center of pressure (CP) at each digit was computed as
 | (1) |
 | (2) |
ANOVAs were used to determine the effect of weight location and rotation (independent variables) on movement variables and force amplitude relative to baseline (dependent variable). A significance level of 0.05 was used for all analyses. All values are mean ± SE unless otherwise noted.
The EMG signal for each muscle was rectified and then low-pass filtered at 20 Hz (d'Avella et al. 2006) prior to analysis. Mean EMG traces for each muscle were computed across trials for each rotation in each condition. These mean EMG waveforms were then normalized to the maximum amplitude observed for that muscle during the entire experiment.
Principal component analysis
Principal component (PC) analysis was used to characterize the patterns of forces utilized for producing object rotations for each object configuration, i.e., weight up and weight down. A PC analysis was performed using the Matlab "princomp" function as described in Klein Breteler et al. (2007). The inputs to the PC analysis of force were the contact forces (Fx, Fy, and Fz), from 25% before rotation onset to 25% after rotation offset (from individual subjects and averages over all subjects) for each digit and each rotation. In this analysis, we used the covariance matrix of 54 force traces (3 digits x 3 forces x 6 rotations). Each force waveform could be perfectly reconstructed as the average force magnitude of each contact force for each dimension (Fx, Fy, and Fz) plus a weighted sum of the 54 PCs. Cross-correlations of weighting coefficients were performed to illustrate which forces tended to change together. PC analysis revealed that PC1 accounted for >85% of the variance and PC2 accounted for 13% of the variance. PC3 never contributed >1%, therefore only PC1 and PC2 were included in the cross-correlation analysis.
PC analysis was also performed on EMG data,; however, the input for the EMG PC analysis was the mean amplitude normalized EMG signal, over the same time interval, computed across weight-up and -down object configurations, for each subject. The PC analysis was performed on individual subjects because EMGs tended to be more variable across subjects. The analysis used the covariance matrix from 60 EMG traces (10 muscles x 6 rotations). Each EMG trace could be perfectly reconstructed as the average EMG magnitude of each muscle plus a weighted sum of the 60 PCs. To examine patterns across subjects, the mean PCs as well as the weighting coefficients were computed. Cross-correlations of the mean weighting coefficients were performed to illustrate which muscles tended to change together. Similar to the force PC analysis, PC1 accounted for >80% of the variance, PC2 accounted for <10% of the variance, and PC3 never contributed >3% of the variance. Therefore only PC1 and PC2 were included in the cross-correlation analysis.
|
|
RESULTS |
|---|
|
|
The force patterns reflect changes due to gravitational and angular acceleration components during the rotations. For example, during pitch rotations, the load is transferred from Fz at all contacts to Fy at one or two contacts resulting in a relatively slow monotonic increase in Fy at the respective digits (Fig. 2). For forward (positive pitch) rotations, the load is transferred to the thumb, which is reflected as a large increase in amplitude of FTy during the rotation. Conversely, during backward rotations (negative pitch) the load is transferred to the ring and index finger contacts (FRy and FIy). The rotations also required that the contact forces produce a force couple. When the CM was in the plane of the contact points (weight up), the inertial torque due to angular acceleration was small (0.005 N·m). In the weight-down condition, the inertial torque for pitch and roll was larger (0.015 N·m). By comparison, the static gravitational torque in the weight-down condition for pitch and roll was an order of magnitude larger (0.09 N·m). Assuming a contact force of 5 N, a change in its moment arm on the order of 1–3 mm could produce inertial torques for a rotation. By contrast, for the weight-down condition, balancing the gravitational torque would require a substantially larger change in moment arm (18 mm).
Force patterns during roll rotations exhibited similar patterns with respect to changes in load distribution and modulation due to angular acceleration. In roll rotations, the load was transferred to the thumb and one digit, either the ring or index finger during left or right rotations, respectively. This was reflected by expected changes in the tangential force at the thumb contact and changes in the normal force at the ring or index finger. Although the load on the thumb after rotation was tangential to the contact surface (FTx), increases in thumb normal forces were also observed during roll rotations. The average (±SD) increases in thumb normal force during left and right rotations were 1.99 ± 1.63 N and 3.96 ± 2.42 for the weight-up condition and 5.44 ± 1.90 N and 7.88 ± 2.17 N for the weight-down condition.
For yaw rotations, where the load direction remains the same, changes in tangential force (<1.0 N) were observed at each digit, consistent with the direction of rotation. Monotonic increases in normal force at each contact that peaked after 50% of the rotation were also observed during yaw rotations. Average (±SD) increases in normal force during yaw rotations were 2.00 ± 0.24 N for the thumb, 1.68 ± 0.70 N for the ring finger, and 1.26 ± 0.12 N for the index finger. One exception to this observation was FRy during counterclockwise rotations for the weight up condition, which peaked around 25% of the rotation. With the preceding exception, the temporal patterns of forces were similar across conditions. However, force amplitude was significantly greater across all rotations when the center of mass was shifted away from the contact plane, i.e., weight-down condition (P < 0.05; Fig. 2) with the exception of FTx, FRz, and FIz.
Results of the PC analysis captured these observations of force patterns. For the weight-up and -down conditions, PC1 accounted for 86 and 95% of the variance, whereas PC2 accounted for 12 and 4% of the variance, and PC3 accounted for 1 and 0.7% of the variance, respectively. The time course of PC1 was very similar to the time course of rotation with a temporal offset, PC1 leading rotation by 100 ms (cf. Fig. 2). This was also indicated by large R2 values from regressions of PC1 onto rotations for individual subjects (mean R2 = 0.87, range: 0.39–1.00) and averaged data (mean R2 = 0.89, range: 0.85–0.92). For weight-up and -down conditions, PC1 had a similar temporal pattern, namely a gradual monotonic change in amplitude, with a larger amplitude for the weight down condition (Fig. 3A). Thus the monotonic increase observed in PC1 is associated with the gravitational component of the rotation. The weighting of each force variable onto PC1 is shown in Fig. 3B for the weight-down condition. Positive weighting of this component for normal forces at each contact for all rotations reflects the tendency for normal forces to increase during the rotation, even for yaw rotations when load did not change (Fig. 3B, top). Furthermore, larger weightings occur when the load was transferred to that contact, for example, the thumb normal force has a larger weighting on PC1 for the forward compared with backward rotation. Paired t-test revealed FTy weights were 95% higher for forward versus backward rotations (df = 18, P < 0.05), FRy weights were 81% higher for left versus right rotations (df = 18, P < 0.05), and FIy weights were 83% higher for right versus left rotations (df = 18, P < 0.05) for the weight-up condition. Smaller nonsignificant differences ranging from 38 to 73% were observed for the weight-down condition. Relatively large negative PC1 weightings for FTz, FRz, and FIz for the forward, left, and right rotations, respectively, denote rotations where there was an obligatory sign change at that contact (see Fig. 2, Tz, forward rotation).
|
; Winges et al. 2007). This component was substantial for Fy and Fz for pitch and roll rotations (Fig. 3B), and these forces could contribute to pitch and roll rotations by producing moments. Regression analysis was performed to examine the extent to which PC2 was associated with angular velocity. Mean R2 values from regressions of PC2 onto angular velocity were 0.79 (range: 0.22–0.99) for individual subjects and 0.90 (range: 0.83–0.96) for averaged data. The third PC, which was consistently small, resembled the angular acceleration profile, reflecting the inertial component of the motion. Regression analysis of PC3 onto angular acceleration revealed a weak relation with mean R2 values of 0.34 (range: 0.02–0.76) for individual subjects data and 0.45 (range: 0.33–0.59) for averaged data. EMG
We recorded EMG activity from five extrinsic and five intrinsic hand and wrist muscles during the rotations (Fig. 4). Although the magnitude of muscle activity for a given muscle varied across subjects for a single rotation, consistent temporal patterns of muscle activity were observed across subjects and conditions. In the following, we will describe the patterns of muscle activity that were common across all subjects. For each rotation, specific sets of muscles tended to have activity characterized by monotonic increases around the onset of the rotation to a peak near 50% of the rotation, followed by a gradual decay to a level higher than that at the onset of rotation. This pattern of activity was exemplified by ED and ECU during forward rotations. Other muscles such as FCU (FWD rotation) and LUMr (BWD and LFT rotation) exhibited a sustained increase in activity. Rotation specific muscle activation patterns were observed for all extrinsic hand muscles we recorded (Fig. 4, 1st–5th rows). For pitch rotations, ED and ECU were activated during forward rotations but showed only small changes in activity during backward rotations; conversely, activity in FD, FD2, and FCU increased by a small amount during forward rotations with much more marked activity during the opposite backward rotation (Fig. 4, 1st and 2nd columns). Thus these sets of muscles appear to be antagonistic for pitch rotations, but this pattern is not maintained for roll and yaw rotations. For example, during roll rotations to the right and yaw clockwise rotations, ECU was activated in concert with FD, FD2, and FCU.
|
Therefore extrinsic as well as intrinsic hand muscles are actively controlled in a manner that is specific to the direction of the rotation being performed.
Because the patterns of EMG activity tended to be similar for the two conditions (weight up and weight down), the PC analysis was performed on EMG averaged across conditions. The resulting PCs for EMG were similar to those obtained in the force PC analysis (Fig. 3A) and were consistent with our observations of the EMG activation patterns. PC1, which accounted for 80% of the variance, reflects the monotonic increase to a peak, followed by a gradual decline described in the preceding text. PC2, which accounted for 9% of the variance, was phasic, with an early onset of EMG activity peaking after peak acceleration. The weighting of each muscle's EMG onto PC1 and PC2 is shown in Fig. 3C. For pitch rotations, the weighting coefficients of PC1 were largest for ED and ECU in the forward direction but were near zero in the backward direction, whereas the opposite pattern was seen in FD, FD2, and FCU. The FDI was the only intrinsic muscle with a relatively large difference in weighting across pitch rotations, i.e., nearly zero for forward rotations and the largest weighting among intrinsic muscles for backward rotations (Fig. 3C, top). For roll rotations, FD, FD2, and FCU had small weightings for the left rotations that increased threefold for the right rotations, whereas the opposite was observed for ECU. APB, FPB, and LUMr had relatively large weightings that were consistent across roll rotations, whereas FDI and LUMi weightings were small but larger for right compared with left rotations. For the yaw rotations, ECU, followed by FCU and LUMr, had the largest weighting coefficients for clockwise rotations but were near zero for the counterclockwise rotations. The opposite was observed for APB, FPB, and FDI (Fig. 3C, top).
PC2 weightings were positive and relatively large for LUMr during both yaw rotations and ECU during clockwise rotations; this is consistent with the early onset of EMG activity observed for these muscles (Fig. 4). Relatively large negative PC2 weighting of APB and FPB for the left rotations is consistent with the late onset of phasic EMG observed during this rotation. In summary, the weighting of the first two PCs for EMG data described above captured the overall EMG patterns (Fig. 3), suggesting that two basic rotation specific EMG patterns can account for the control of hand-held object rotation.
Force and EMG patterns
To examine the extent to which patterns of force (in different digits) and EMG modulation (in different muscles) were consistent across rotations, we performed cross-correlations on the PC1 and PC2 weighting coefficients for pairs of force components and EMGs (Fig. 5). We computed a cross-correlation matrix and reordered the columns and rows of this matrix to show highly correlated variables next to each other. The warm colors (i.e., red) indicate positive correlations between weightings while the cool colors (i.e., blue) indicate negative correlation. A strong positive correlation between weights from two or more input variables may indicate a synergistic relation between the respective pairs of forces or muscles.
|
The strength of correlations across force variables for PC2 was similar to PC1 although the correlations were stronger when RZ was paired with IY or TY. Correlations across EMG PC2 weights also had similar patterns to PC1, although the strength of correlations between LUMi and APB or FPB was stronger, whereas correlations between LUMi and FD2 or LUMr were weaker. Because PC2 is associated with early onset of EMG, strong relations between intrinsic thumb and index finger muscles suggest that intrinsic thumb and index muscles tend to have similar onset times. The finding of correlations across both extrinsic and intrinsic muscles is important because it shows active control of the intrinsic muscles with respect to the rotation movement, the alternative being co-contraction among all intrinsic hand muscles to maintain a stiffened hand during the rotation.
CP and force direction
As noted in the INTRODUCTION, a moment of force could be produced by shifting the CP at one or more digits. CP changes in the horizontal x dimension were largest during yaw rotations and were consistently in the direction of the rotation at all three contacts (Fig. 6, bottom), whereas changes in the z direction tended to be larger for pitch and roll rotations. There was no difference across conditions (weight up and weight down) in the magnitude of CP changes for CPx or CPz at any contact (P > 0.05). Overall, mean (±SD) changes in CP were small: 2.00 ± 0.91 and 2.39 ± 0.31 mm for the x and z dimensions, respectively. At the thumb contact, CPx changes were largest for clockwise rotations (4.42 ± 0.22 mm; P < 0.001) whereas CPz changes were largest for forward and backward rotations (2.95 ± 0.23 and 2.65 ± 0.23 mm, respectively; P < 0.001). At the ring finger contact, CPx changes were larger for clockwise compared with right rotations (3.10 ± 0.31 and 1.54 ± 0.31 mm, respectively; P < 0.05), whereas CPz changes were largest for left rotations (4.25 ± 0.32; P < 0.001). CPx changes at the index finger contact were largest for clockwise and counterclockwise rotations (3.08 ± 0.24 and 2.43 ± 0.24 mm, respectively; P < 0.001) whereas there was no difference in CPz changes across rotations.
|
In addition to changes in force magnitude and CPs at individual contact points, changes in the direction of force vectors could contribute to the moment required to produce a rotation of a hand-held object. The force vector within the grip plane was computed for each digit to examine how force direction changes with respect to rotation and condition. The largest changes in force vector direction tended to occur 25% of normalized movement time (Fig. 6, yellow arrows). Furthermore, the results show that the thumb force vector was directed close to perpendicularly to the contact surface (pitch and yaw) or it was directed toward the finger contact to which the load was transferred during roll rotations (Fig. 6, middle). In the latter case, it is noteworthy that this change in the direction of thumb force anticipates the rotation by 50–100 ms in all subjects. During the rotation, the index force vector exhibited the largest direction changes (especially for yaw rotations), whereas the ring force vector exhibited the smallest direction changes that were similar across all rotations.
The mean (±SD) change in force direction was 7.72 ± 2.67° and ranged from 3.62 to 16.45° across all digits, rotations, and conditions. The change in thumb force direction was slightly larger for the weight-up versus -down condition (8.64 and 6.38 ± 0.50°, respectively; P < 0.002), whereas there were no differences across condition for the ring or index force vectors. Figure 6 shows force vectors for each digit from the weight-down condition to illustrate changes in force vector magnitude and direction during the rotations. The change in thumb force direction differed across rotation such that larger changes were observed for the backward and left rotations compared with the forward, clockwise, and counterclockwise rotations (9.83 and 11.58 ± 0.86 vs. 3.64, 6.02 and 6.04 ± 0.86°, respectively; P < 0.001; Fig. 6). The right rotation also had a larger thumb force direction change compared with the forward rotation (7.96 and 3.64 ± 0.86°, respectively; Fig. 6, A and B). There were no differences across rotations for the change in ring force direction (P > 0.05). For the index force vector, direction changes were larger for clockwise and counterclockwise rotations than forward and right rotations (14.62 and 14.47 ± 1.35 vs. 6.98 and 6.33 ± 1.35°, respectively; Fig. 6), and larger for left compared with right rotations (12.40 and 6.33 ± 1.35°, respectively; P < 0.001; Fig. 6B).
Together these results indicate that at the thumb and index, changes in CP and force direction are both utilized to varying degrees for specific rotations. However, as we argue in the DISCUSSION, these changes in CP and force direction are due largely to the interaction of a stiff hand with a rigid object rather than active control.
|
|
DISCUSSION |
|---|
|
The modulation in grasp forces as well as in the muscle activity of extrinsic and intrinsic hand muscles during object rotation reflects a very simple temporal pattern of control because two principal components were able to account for >80% of the variance. For most rotations, the first PC in force as well as in EMG reflected changes in gravitational load because it was largely similar to the time course of rotation. The second PC, which was phasic, had a time course that was similar to the angular velocity. The description of forces and muscle activity using only two PCs is indicative of synergistic control. Scaling of the amplitude of muscle patterns related to the rotational position and to the angular velocity could account for the modulation of muscle activity observed during hand-held object rotations. Furthermore, rotation specific patterns of muscle activation were expected for extrinsic hand muscles because they contribute to wrist motion; however, these patterns were also observed for intrinsic hand muscles. Therefore extrinsic as well as intrinsic hand muscles are actively controlled with respect to the rotation being performed. The highly negative correlation between the intrinsic thumb muscles and the extrinsic finger muscles and ring lumbrical also supports the idea of active control in the intrinsic muscles.
Force patterns
Rotation of a hand-held object can be produced using a variety of strategies. However, across all subjects there were some distinct patterns that emerged for rotations about each axis. For changes in pitch, the thumb and finger force vectors opposed one another. Similar to static grasping, the thumb force vector was typically directed between the two fingers, which is consistent with the thumb opposing a virtual finger created by the resultant of the finger force vectors (Baud-Bovy and Soechting 2001). However, for changes in roll, the thumb tended to oppose only one digit, i.e., the one that would share in supporting the object's weight at the maximum rotation. For pitch and roll rotations, as force amplitude increased during the rotation, there was very little shift in the CP and the largest changes in force direction tended to occur around 25% of the normalized movement time (see Fig. 6, A and B, yellow arrows). This general strategy was consistent across the weight-up and -down conditions, although the amplitude of the force change was smaller for the weight-up condition.
A different strategy was utilized to produce yaw rotations. Because the object was rotated about the vertical axis, there are no changes in load force thus changes in grip force would not be required to maintain a stable grasp (Burstedt et al. 1999). Nevertheless during the rotation, the magnitude of the force vectors did increase for both weight-up and -down conditions although not to the extent observed for pitch and roll rotations. For yaw rotations, the force direction changed very little at the thumb and ring finger, whereas the CP at each contact did shift to a large extent during the rotations. For clockwise rotations (negative yaw), the CP shifted in the negative direction, whereas the opposite was observed for counterclockwise (positive yaw) rotations. These shifts in CP occurred at all digits and are consistent with rolling of each digit over the surface. This suggests that these changes are passive in nature.
Thus simple force strategies were observed for each rotation. With respect to the grasp plane, forces increased during the rotation whether it was required due to changes in load (for pitch and roll rotations) or not (for yaw rotations). While the CP and force direction also changed at each digit for all rotations, the extent of these changes could be very small. Furthermore changes in CPs did not always go hand in hand with changes in force direction, rather each rotation tended to be dominated by one or the other. Finally, the observed force patterns were consistent with grasp force being the reflection of a neuromuscular strategy as well as a mechanical interaction between the object and a stiffened hand (Winges et al. 2007). Forces related to acceleration (PC3) were small even though the rotation movements were fast, and rotation specific changes in forces in the contact plane reflected primarily the monotonic shifts in position and/or gravitational load (PC1) and angular velocity (PC2).
Modulations in grip force with load force have been described previously (e.g., Flanagan and Wing 1993; Witney and Wolpert 2007). For example, Zatsiorsky et al. (2005) observed that grip force modulations during vertical cyclic arm movements accounted for changes in load force during the movements. Our results are also relevant to the theoretical studies on robotic grasp control by Arimoto and colleagues (2000, 2001), subsequently extended to human experimental work by Zatsiorsky and colleagues (2004). These authors suggested that the forces generated at the contact surfaces result from the superposition of two commands: one controlling grasping strength and the other controlling object position. Our analysis instead subdivides the forces that rotate and hold the object into those actively controlled by muscle contraction (as reflected in the 1st 2 PCs), those produced by passive interaction between a compliant hand and a rigid object and those due to passive (rolling) changes in the CP.
Patterns of muscle activity
A previous study found that co-contraction and modulation of hand muscle activity contributed to the maintenance of a stable grasp during translation movements by adjusting the stiffness at one or more digits (Winges et al. 2007). Similar to horizontal translation movements, the rotation task could have been completed by co-contracting the intrinsic thumb and finger muscles to maintain a stiffened hand, while wrist and arm muscles produced motion at the wrist. However, EMG recorded from wrist, extrinsic, and intrinsic finger muscles demonstrated a rotation specific pattern for each subject. PC analysis confirmed this observation and revealed that a PC with a slow monotonic increase during the rotation could account for 
80% of the variance. The time course of this PC was similar to that revealed in the PC analysis of force. Cross-correlation of PC weights for all muscles and rotations revealed synergistic relations across extrinsic and intrinsic finger and wrist flexor muscles across all subjects. Although their task was different, this finding is consistent with the observations of Werremeyer and Cole (1997), who demonstrated parallel changes in the activity of a primary wrist flexor, i.e., flexor carpi radialis, and FDS and FDI, which are extrinsic and intrinsic finger muscles, respectively, during a rapid wrist flexion movement.
If intrinsic muscles are primarily responsible for maintaining a stiffened hand, a relatively strong correlation among all intrinsic muscles might be expected. However, intrinsic finger muscles were more weakly correlated to intrinsic thumb muscles than to other intrinsic finger muscles. This is consistent with force strategies we observed such that for pitch rotations, the thumb opposed the fingers as a unit or a virtual finger, while in roll rotations the thumb opposed only one of the fingers. Therefore these results indicate that intrinsic as well as extrinsic hand muscles are actively controlled during the rotation of a hand-held object.
Although modulation of muscle activity was typically consistent with changes in force, there was no difference in the amplitude of EMG across conditions, whereas this difference was very apparent in the force data. Therefore the modulation of hand muscle activity may be more closely associated with maintaining stiffness at a particular digit to compensate for additional loads as in the case of pitch and roll rotations. This is consistent with observations from a hand-held object translation task where co-contraction and modulation of muscle activity compensated for inertial forces by controlling the stiffness of the hand (Winges et al. 2007).
The simplicity of control of hand muscle activity during hand-held object rotation was evidenced by two PCs accounting for >85% of the variance. In other tasks, previous studies have required three to five PCs to account for a similar amount of variance in muscle activity. Four or more PCs (or muscle synergies) were required to account for 80% of the variance in EMG patterns for dynamic finger spelling (Klein Breteler et al. 2007) and fast reaching movements (d'Avella et al. 2006), whereas static finger spelling and object grasp (without lift) required three to four PCs (Weiss and Flanders 2004) to account for a similar amount of variance in static EMG levels. The results of the current study indicate that two PCs were capable of describing the muscle activation patterns to a large extent during object rotation. In contrast, previous studies have found that multiple muscle synergies are needed to account for the temporal shift in activation of pairs of asynchronously active muscles (d'Avella et al. 2006; Flanders 1991; Flanders et al. 1996; Klein Breteler et al. 2007). Therefore the main difference is that for our task, groups of muscles can be controlled through amplitude scaling of a simple command without a temporal shift, whereas other tasks require additional commands to account for the temporal shift in activation patterns of groups of simultaneously active muscles. Thus our results suggest that a very simple control strategy is utilized for the modulation of wrist and hand muscles during unidirectional object rotations.
One explanation for the simple control strategy we observed is that the muscles are being controlled primarily to counteract the changes due to gravity. This is similar to the static component of EMG found during vertical plane reaching movements, such that the EMG amplitude changes reflected the effect of gravity at different positions in the vertical plane (Flanders et al. 1996). Additionally, Flanders and Hermann (1992) found that two PCs could describe muscle activation patterns during point-to-point movements in the frontal and sagittal plane.They also found that their two PCs could be transformed into tonic and phasic components that could correspond to the forces required to counteract gravity and those that would produce increases in velocity to produce the movement. Thus it seems reasonable that if the hand maintains a rigid posture during unidirectional object rotations, the largest factor governing muscle control could be described by one component concerned with changes due to gravity. Thus a relatively simple control strategy would be sufficient because it would not require the inclusion of asynchronous muscle activations, something that is an important factor for fast linear movements.
Postural muscle synergies
Other tasks such as those related to posture and balance control have also demonstrated that muscle activation patterns can be reduced down to a few useful control components that are capable of describing muscle activation patterns. During walking, Ivanenko et al. (2004) found that five components could describe patterns of activation for 
20 muscles during walking. Krishnamoorthy et al. (2003a,b) found that postural adjustments due to shifts in CP during standing could be described by only three PCs. Control of balance has also been studied in cats, and it has been revealed that a small number of muscle synergies are used to control balance when perturbations are made from many different directions (Ting and Macpherson 2005). Thus many tasks that are performed by controlling a large number of muscles and joints can be explained using a few simplified patterns of muscle activity, suggesting that synergistic groups of muscles are utilized to simplify otherwise highly complex control requirements.
Research has shown that for kinematic and kinetic variables associated with grasping, a relatively small number of coordination patterns emerge among digits of the hand (for review, see Schieber and Santello 2004). Synergistic activation of muscles and thus production of forces appears to be a consistent component of control for tasks that require a control strategy that encompasses a large number of degrees of freedom. One possible mechanism underlying these coordinated muscle activation and force production patterns is a common neural signal delivered to pools of motor neurons innervating different muscles (i.e., across-muscle synchrony). Modeling has indicated that a moderate level of across-muscle synchrony could contribute to the coordination of hand muscles and the forces they produce (Santello and Fuglevand 2004). Across-muscle synchrony has been observed across extrinsic and intrinsic hand muscles with varying strengths during grasping (Hockensmith et al. 2005; Keen and Fuglevand 2004; McIsaac and Fuglevand 2006; Winges and Santello 2004; Winges et al. 2006, 2008). Although there is still some question as to the extent to which the strength of this common neural input is modified with respect to task, it seems there may be a fixed gradient among hand muscles during static grasping tasks such that common neural input is strongest across extrinsic hand muscles and weakest across combinations of intrinsic hand muscles (Winges et al. 2008). This type control signal could be one of the mechanisms underlying the synergistic control across hand muscles that would allow for the use of a relatively simple control strategy for tasks involving object grasping and manipulation.
|
|
GRANTS |
|---|
|
|
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: S. A. Winges, 6-145 Jackson Hall, 321 Church St. SE, Minneapolis, MN, 55455 (E-mail: swinges{at}umn.edu)
|
|
REFERENCES |
|---|
|
Arimoto S, Tahara K, Yamaguchi M, Nguyen PTA, Han H-Y. Principles of superposition for controlling pinch motions by means of robot fingers with soft tips. Robotica 19: 21–28, 2001.[CrossRef][Web of Science]
Baud-Bovy G, Soechting JF. Two virtual fingers in the control of the tripod grasp. J Neurophysiol 86: 604–615, 2001.
Burstedt MK, Flanagan JR, Johansson RS. Control of grasp stability in humans under different frictional conditions during multidigit manipulation. J Neurophysiol 82: 2393–2405, 1999.
d'Avella A, Portone A, Fernandez L, Lacquaniti F. Control of fast-reaching movements by muscle synergy combinations. J Neurosci 26: 7791–7810, 2006.
Flanagan JR, Wing AM. Modulation of grip force with load force during point-to-point arm movements. Exp Brain Res 95: 131–143, 1993.[Web of Science][Medline]
Flanders M. Temporal patterns of muscle activation for arm movements in three-dimensional space. J Neurosci 11: 2680–2693, 1991.[Abstract]
Flanders M, Herrmann U. Two components of muscle activation: scaling with the speed of arm movement. J Neurophysiol 67: 931–943, 1992.
Flanders M, Pellegrini JJ, Geisler SD. Basic features of phasic activation for reaching in vertical planes. Exp Brain Res 110: 67–79, 1996.[Web of Science][Medline]
Goodwin AW, Jenmalm P, Johansson RS. Control of grip force when tilting objects: effect of curvature of grasped surfaces and applied tangential torque. J Neurosci 18: 10724–10734, 1998.
Hockensmith GB, Lowell SY, Fuglevand AJ. Common input across motor nuclei mediating precision grip in humans. J Neurosci 25: 4560–4564, 2005.
Ivanenko YP, Poppele RE, Lacquaniti F. Five basic muscle activation patterns account for muscle activity during human locomotion. J Physiol 556: 267–282, 2004.
Johansson RS, Backlin JL, Burstedt MK. Control of grasp stability during pronation and supination movements. Exp Brain Res 128: 20–30, 1999.[CrossRef][Web of Science][Medline]
Keen DA, Fuglevand AJ. Common input to motor neurons innervating the same and different compartments of the human extensor digitorum muscle. J Neurophysiol 91: 57–62, 2004.
Klein Breteler MD, Simura KJ, Flanders M. Timing of muscle activation in a hand movement sequence. Cereb Cortex 17: 803–815, 2007.
Krishnamoorthy V, Goodman S, Zatsiorsky V, Latash ML. Muscle synergies during shifts of the center of pressure by standing persons: identification of muscle modes. Biol Cybern 89: 152–161, 2003a.[CrossRef][Web of Science][Medline]
Krishnamoorthy V, Latash ML, Scholz JP, Zatsiorsky VM. Muscle synergies during shifts of the center of pressure by standing persons. Exp Brain Res 152: 281–292, 2003b.[CrossRef][Web of Science][Medline]
McIsaac TL, Fuglevand AJ. Influence of tactile afferents on the coordination of muscles during a simulated precision grip. Exp Brain Res 174: 769–774, 2006.[CrossRef][Web of Science][Medline]
Santello M, Fuglevand AJ. Role of across-muscle motor unit synchrony for the coordination of forces. Exp Brain Res 159: 501–508, 2004.[CrossRef][Web of Science][Medline]
Schieber MH, Santello M. Hand function: peripheral and central constraints on performance. J Appl Physiol 96: 2293–2300, 2004.
Smith MA, Soechting JF. Modulation of grasping forces during object transport. J Neurophysiol 93: 137–145, 2005.
Ting LH, Macpherson JM. A limited set of muscle synergies for force control during a postural task. J Neurophysiol 93: 609–613, 2005.
Weiss EJ, Flanders M. Muscular and postural synergies of the human hand. J Neurophysiol 92: 523–535, 2004.
Werremeyer MM, Cole KJ. Wrist action affects precision grip force. J Neurophysiol 78: 271–280, 1997.
Winges SA, Johnston JA, Santello M. Muscle-pair specific distribution and grip-type modulation of neural common input to extrinsic digit flexors. J Neurophysiol 96: 1258–1266, 2006.
Winges SA, Kornatz KW, Santello M. Common input to motor units of intrinsic and extrinsic hand muscles during two-digit object hold. J Neurophysiol (Jan 2, 2008). doi:10.1152/jn.01059.2007.
Winges SA, Santello M. Common input to motor units of digit flexors during multi-digit grasping. J Neurophysiol 92: 3210–3220, 2004.
Winges SA, Soechting JF, Flanders M. Multidigit control of contact forces during transport of handheld objects. J Neurophysiol 98: 851–860, 2007.
Witney AG, Wolpert DM. The effect of externally generated loading on predictive grip force modulation. Neurosci Lett 414: 10–15, 2007.[CrossRef][Web of Science][Medline]
Yoshikawa T, Nagai K. Manipulating and grasping forces in manipulation by multifingered robot hands. IEEE Trans Robotics Automation 7: 67–77, 1991.[CrossRef]
Zatsiorsky VM, Gao F, Latash ML. Motor control goes beyond physics: differential effects of gravity and inertia on finger forces during manipulation of hand-held objects. Exp Brain Res 162: 300–308, 2005.[CrossRef][Web of Science][Medline]
Zatsiorsky VM, Latash ML, Gao F, Shim JK. The principle of superposition in human prehension. Robotica 22: 231–234, 2004.[CrossRef][Web of Science]
This article has been cited by other articles:
|
|
 |
|
  S. A. Winges, S. E. Eonta, J. F. Soechting, and M. Flanders Effects of Object Compliance on Three-Digit Grasping J Neurophysiol, May 1, 2009; 101(5): 2447 - 2458. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
TOP
ABSTRACT
INTRODUCTION
