| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139; and 2 Vision et Motricité, Institut National de la Santé et de la Recherche Médicale Unité 94, 69500 Bron, France
 |
ABSTRACT |
|---|
 Abstract
 Introduction
Methods
Results
Discussion
References
|
|---|
Desmurget, Michel and Claude Prablanc. Postural control of three-dimensional prehension movements. J. Neurophysiol. 77: 452-464, 1997. This experiment was carried out to test the hypothesis that three-dimensional upper limb movements could be initiated and controlled in the joint space via a mechanism comparing an estimate of the current postural state of the upper arm with a target value determined by one specific inverse static transform converting the coordinates of the object into a set of arm, forearm, and wrist angles. This hypothesis involves two main predictions: 1) despite joint redundancy, the posture reached by the upper limb should be invariant for a given context; and 2) a movement programmed in joint space should exhibit invariant characteristics of the joint covariation pattern as well as a corresponding variable hand path curvature in the task space. To test these predictions, we examined prehension movements toward a cylindrical object presented at a fixed spatial location and at various orientations without vision of the moving limb. Once presented, the object orientation was either kept constant (unperturbed trials) or suddenly modified at movement onset (perturbed trials). Three-dimensional movement trajectories were analyzed in both joint and task spaces. For the unperturbed trials, the task space analysis showed a variable hand path curvature depending on object orientation. The joint space analysis showed that the seven final angles characterizing the upper limb posture at hand-to-object contact varied monotonically with object orientation. At a dynamic level, movement onset and end were nearly identical for all joints. Moreover, for all joints having a monotonic variation, maximum velocity occurred almost simultaneously. For the elbow, the only joint presenting a reversal, the reversal was synchronized with the time to peak velocity of the other joint angles. For the perturbed trials, a smooth and complete compensation of the movement trajectory was observed in the task space. At a static level the upper limb final posture was identical to that obtained when the object was initially presented at the orientation following the perturbation. This result was particularly remarkable considering the large set of comfortable postures allowed by joint redundancy. At a dynamic level, the joints' covariation pattern was updated to reach the new target posture. The initial synergies were not disrupted by the perturbation, but smoothly modified, the different joints' movements ending nearly at the same time. Taken together, these results support the hypothesis that prehension movements are initiated and controlled in the joint space on the basis of a joint angular error vector rather than a spatial error vector.
Since the initial work of Woodworth (1899) Subjects
Six right-handed subjects (5 males, 1 female) from 22 to 48 yr of age participated in this experiment. All of the subjects were naive with regard to the task and the purpose of the experiment.
Experimental procedure
A schematic representation of the apparatus is presented in Fig. 1. The subject was seated comfortably in a dentist's chair. The subject's trunk was immobilized by a harness to prevent any displacement of the shoulder axis during the task. In front of the subject, a fast servo-controlled torque motor supported the object to be grasped, i.e., a cylinder having a weight of 400 g, a diameter of 5 cm, and a grip length of 10 cm. The object's center of mass was located in the parasagittal plane crossing the shoulder of the subject. The distance between object and shoulder was equal to 80% of the upper limb length of the subject. The motor, whose axis of rotation was horizontal and sagittally oriented, allowed the object to be tilted in the frontoparallel plane. Because the whole experiment was carried out in a dark room, the object was equipped with an electronic device allowing illumination from the inside. Another electronic device, also located inside the object, allowed the nearly instantaneous detection of contact between the hand and the object (delay <10 ms).
Recording technique
The experiment was controlled on-line by a program running on an IBM PC486. Movements were bidimensionally recorded at a frequency of 208 Hz by a SELSPOT II system equipped with two cameras. A direct linear transform method was used to reconstruct the three-dimensional coordinates of six infrared-emitting diodes placed on the right arm of the subject in the following positions: 1) metacarpophalangeal joint of the index finger; 2) metacarpophalangeal joint of the auricular finger; 3) radial styloid; 4) cubital styloid; 5) above the cubital head of the elbow; and 6) external extremity of the accromion. For each diode, the X, Y, and Z position data were filtered at 10 Hz with a zero-phase finite impulse response filter, with the use of 33 coefficients.
Spatial analysis
The wrist position, defined as the center of gravity of the infrared-emitting diodes 3 and 4, was used to reconstruct the three-dimensional trajectory of the movements (Jeannerod 1981
Joint analysis
The following seven angles were used to determine the upper limb posture at each time step (Fig. 3); the angular reconstruction method has been described in a preceding paper (Desmurget et al. 1995a Blocked versus control trials
As demonstrated by the ANOVA 1, the interactions between orientation (5 levels: Early versus late perturbed trials
As shown by the ANOVA 2, the interactions between the orientation (5 levels: Control versus perturbed trials
SPATIAL ANALYSIS
Wrist path divergence among unperturbed trials. For all the orientations of the object, the three-dimensional path of the wrist exhibited a general curved shape (Fig. 4). The values at which the C60, C40, C0, and C
KINEMATIC ANALYSIS.
The most interesting observation related to the kinematic analysis concerned the morphological similarity observed between the perturbed and unperturbed movements for both the velocity and acceleration profiles. Strikingly, the perturbed curves presented no secondary peaks that could be related to the modification of the object orientation at movement onset (Fig. 6). This result, which showed that the initial motion was not braked and interrupted to allow the initiation of a corrective movement, suggested strongly that the trajectory corrections were performed by updating the current motor command. Note that the pattern of the individual curves was the same as that of the averaged curves, demonstrating that the absence of secondary peaks on averaged trajectories was not a computational artifact induced by the averaging procedure (Fig. 6).
POSTURAL ANALYSIS
Static analysis. Table 1 shows the mean ± SD (n = 6 subjects) for the seven upper limb angles defining the final arm posture. The same data are displayed in Fig. 7 for a representative subject.
In summary, the present experiment provides a large set of converging arguments in support of the hypothesis that the final posture to be reached plays a key role in movement control by defining an internal reference to which the current state of the system is continuously compared. The overall view emerging from this study agrees with the following description. The spatial characteristics of the object to grasp are initially converted into a set of arm and forearm orientations. The movement from the starting posture to the target posture is then implemented on the basis of an "angular error vector" whose components represent the difference between the starting and target angles for each joint. During the movement, joint angle variations are not controlled independently, but in a synergic way (temporal coupling). The movement trajectory observed, either in the task or joint space, results directly from this temporal coupling. As shown by the motor responses observed during the perturbed trials, neither the final posture to be reached nor the characteristics of the initial joint synergies is rigidly fixed at movement onset. When the initial response is maladjusted, for instance because of a wrong definition of the final posture to be reached, the motor system is able to update the target posture during the course of the motion and to reorganize the joint synergies accordingly. Both this hypothesis and the concept of postural control will be addressed in more details in the following.
Postural control hypothesis
Because the number of degrees of freedom of the upper limb (df = 7 when finger joints are neglected) exceeds those necessary to completely specify the location and orientation of an object in space (df = 6), the mathematical relationship associating the coordinates of the object to grasp and the final posture of the arm (inverse mapping) is a priori indeterminate (note that the object's df was not equal to 6 but to 5 in the present experiment, because the object was cylindrical). Regarding the task studied in the present experiment, this observation indicates that a final given orientation of the hand in the frontoparallel plane can be theoretically obtained through an infinite number of angular combinations. For example, a given subject can adjust hand orientation to object tilt with the use of mainly forearm rotation or elbow elevation. The subject can also combine these two degrees of freedom. Interestingly, our data showed that all the available degrees of freedom of the upper limb were used and combined by the nervous system to adjust hand orientation. As demonstrated by the static postural analysis, joint redundancy was not eliminated by blocking some degrees of freedom (Feldman and Levin 1995 Postural control and motor equivalence
The ability to correct the movement during its execution is generally associated with the concept of motor equivalence (Abbs and Gracco 1984 Joint synergies during postural transition
The cornerstone of behavioral neuroscience lies in the postulate that spatiotemporal invariances can be used as insights to understand the fundamental principles underlying movement generation (Morasso 1981 Reorganization of ongoing movements
During the perturbed trials, a complete compensation of the movement trajectory was observed. The first sign of kinematic modification was identified on the wrist velocity vector orientation after a delay of 290 ms. This value, which appears relatively high with respect to previous double-step studies cat: Alstermark et al. 1990 (100 ms); humans: Gentillucci et al. 1992 (150 ms); Paulignan et al. 1991
INTRODUCTION
Abstract
Introduction
Methods
Results
Discussion
References
, many psychophysical studies have suggested that goal-directed movements could be segmented into two components: the first one "ballistic" and ensuring only the transport of the hand into the vicinity of the goal, the second one "controlled" and allowing through feedback mechanisms the accurate acquisition of the target (Arbib 1981; Keele 1981; Meyer et al. 1988; see Jeannerod 1988 for a review). The validity of this widespread model has been progressively challenged by pointing experiments showing that the initial stage of visually directed movements could be amended smoothly when target location was displaced at movement onset (Goodale et al. 1986; Pélisson et al. 1986; Prablanc and Martin 1992). To account for this observation, several authors suggested that visually directed movements were continuously controlled during their execution by a feedback loop comparing the respective locations of the hand and the target (Desmurget et al. 1995b; Pélisson et al. 1986; Prablanc and Martin 1992; Prablanc et al. 1986; Redon et al. 1991; Van Sonderen et al. 1989). According to this hypothesis, usually labeled the spatial control hypothesis, the difference between the respective locations of the hand and target is used by the motor system to update the neural command forwarded to the muscles. This model, also proposed to account for the processes underlying phonemic production (Abbs and Gracco 1984) and gaze orientation (Guitton et al. 1990; Laurutis and Robinson 1986; Pélisson et al. 1988), has recently received a computational validation: as demonstrated by Hoff and Arbib (1992, 1993; also see Van Sonderen and Van der Gon 1990), motor control models, including internal predictors to compensate for delays in sensory feedback and processing a spatial motor error, faithfully account for the kinematic characteristics of the arm trajectory corrections observed during two-dimensional double-step experiments.
; Flanders et al. 1992; Rosenbaum et al. 1995). As suggested by these experiments, a valid and economical alternative to the spatial control hypothesis could reside in a postural control hypothesis. On the basis of recent studies showing that postural variables are taken into account by the motor system to plan the movement (Flanders et al. 1992; Helms-Tillery et al. 1991; Hore et al. 1992, 1994; Lacquaniti and Maioli 1994a,b; Miller et al. 1992; Rosenbaum et al. 1990, 1992; Scott and Kalaska 1995; Straumann et al. 1991), this latter hypothesis proposes that the final posture to be reached constitutes the critical parameter controlled by the CNS during visually directed movements. According to this view, which is conceptually affiliated with the equilibrium point hypothesis initially formulated by Feldman (see Feldman 1986 for an overview), the spatial characteristics of the object to grasp are initially converted into a set of arm and forearm orientations. The arm movement is then initiated and controlled in a joint space via a main mechanism continuously comparing an estimate of the current postural state of the arm with the "target value" to be reached.
; Stelmach et al. 1994)]. To test these predictions, we performed a double-step experiment in which subjects were required to grasp a cylindrical object presented at a given spatial location with different orientations. For the perturbed and unperturbed trials, movements were analyzed in both the task and joint space.
METHODS
Abstract
Introduction
Methods
Results
Discussion
References
View larger version (33K):
[in a new window]
FIG. 1.
Schematic representation of the experimental apparatus. The subject sat comfortably on a dentist's chair. The subject's trunk was immobilized by a harness to prevent any movement of the shoulder during the experiment. In front of the chair a fast servo-controlled torque motor supported the object to be grasped. The plate supporting the object was located in the parasagittal plane crossing the shoulder of the subject. The motor allowed the object to rotate in the frontoparallel plane. An electronic circuit instantaneously detecting the onset of hand movement could control, via the motor, a very fast change in object orientation (40° in <50 ms). Another electronic circuit permitted the illumination of the object from inside. The experiment was controlled on-line by a program running on an IBM PC. A Selspot II system equipped with 2 cameras was used to record the displacement of the infrared-emitting diodes placed on the right arm.
View larger version (28K):
[in a new window]
FIG. 2.
Schematic illustration of both the spatiotemporal organization of the task and the main kinematic landmarks used in this experiment. At the beginning of each trial, the right forearm of the subject rested on a tilted plane fixed on the side of the chair (see Fig. 1). During this rest period, the subject fixed a red light-emitting diode placed in front of his body. After the cylinder was moved to 1 of its predetermined orientations, a tone was emitted and the object was lit, indicating that the subject had to grasp it. The whole experiment was performed in the dark, precluding any corrections based on the simultaneous vision of the target and of the moving limb. Two types of trials were considered. During the unperturbed trials, the initial orientation of the object remained constant until movement completion. During perturbed trials, the initial orientation of the object was suddenly and unexpectedly modified at movement onset. For the unperturbed trials, 5 basic orientations ranging from 60° to 
20° were studied. These basic orientations were combined during double-step trials, and 4 perturbed conditions were considered:20° 
60°, 20° 40°, 20° 0°, and20° 20°. Perturbed trials intermixed randomly with unperturbed trials. The main kinematic landmarks studied in this experiment are represented at bottom as follows: TPA, time to peak acceleration; TPV, time to peak velocity; APA, amplitude of peak acceleration; APV, amplitude of peak velocity; DT, deceleration time; MD, hand movement duration; HL, hand latency.
). Each type of movement was repeated 10 times (10 repetitions × 5 orientations = 50 trials). The second session (S2) was identical to the first one, except that the unperturbed trials (here called control trials, designated C) were randomly mixed with 20% perturbed trials (designated P; in other words, 20% of the trials performed during S2 were perturbed trials). Perturbed trials involved the same basic orientations as those used for the unperturbed trials. However, for obvious reasons related to the duration of the experiment, all the possible perturbations involving the five basic orientations of the object could not be studied. Only the combinations for which the initial orientation of the object was 20° were selected and analyzed in the present study. During perturbed trials, the object, whose initial orientation was 20°, was suddenly and unexpectedly tilted to 20° (P20), 0° (P0), 40° (P40), or 60° (P60) at movement onset. Because the number of perturbed trials was fixed at 20% of the total amount of trials to avoid any specific expectation strategies or learning process, each kind of unperturbed trial was repeated 32 times (32 repetitions × 5 orientations = 160 trials), whereas each kind of perturbed trial was repeated 10 times (10 repetitions × 4 orientations = 40 trials). The subject was not informed of the possible occurrence of the perturbations. Moreover, all of the six subjects underwent both S1 and S2, i.e., 250 trials.
; Paulignan et al. 1991). Wrist velocity was computed from the filtered position signal with the use of a least-square second-order polynomial method (window ±4 points). The same method was used to compute the wrist's acceleration from the velocity signal. The main kinematic parameters analyzed in this experiment were (see Fig. 2): hand latency (labeled HL; time between object illumination and movement onset), movement duration (labeled MD; time between movement onset and hand-object contact), time to peak acceleration (labeled TPA), amplitude of peak acceleration (labeled APA), time to peak velocity (labeled TPV; acceleration time), and amplitude of peak velocity (labeled APV).
control analysis), with object orientation (P = 5) and session (q = 2) being the repeated-measures factors. A second ANOVA (ANOVA 2longitudinal analysis) with repeated measures was performed to test whether the early and late perturbed trials differed from each other (learning process). For this ANOVA, the first three (early trials) and last three (late trials) perturbed trials were averaged. Two factors were then considered: the object orientation factor (P = 5) and the learning factor (q = 2: early trials; late trials). A last ANOVA (ANOVA 3perturbation sensitivity) with repeated measures was performed to test the effect of perturbations on movement kinematic landmarks. Nine conditions (repeated-measure factor) were then considered, i.e., four perturbed conditions (P20, P0, P40, P60) and five control conditions (C60, C40, C20, C0, C20). The Duncan multiple-range test was used for post hoc comparisons of the means. Threshold for statistical significance was set at 0.05 (as for all the statistical analyses carried out in this work).
View larger version (50K):
[in a new window]
View larger version (60K):
[in a new window]
View larger version (13K):
[in a new window]
FIG. 3.
Schematic representation of the upper limb angles computed during this experiment. A and B: D1 is defined as being the plane containing the shoulder (S), the wrist (W), and the elbow (E). D2 is defined as being the plane orthogonal to the shoulder-elbow axis. D3 is defined as being the vertical plane containing the shoulder-elbow axis. D4 is defined as being the sagittal plane crossing the shoulder. L1 is the horizontal line contained in D2 and crossing the elbow. L2 is the intercept of the planes D1 and D2. L3 is the horizontal line contained in D3 and crossing the shoulder. Considering the previous definitions, 
(i.e., the angle between L1 and L2) characterized upper arm rotation; 
(i.e., the angle between D1 and the radiocubital axis) characterized forearm rotation; 
(i.e., the angle between D3 and D4) characterized upper arm azimuth; 
(i.e., the angle between L3 and the shoulder-elbow axis) characterized upper arm elevation; and 
(i.e., the angle between shoulder-elbow and elbow-wrist axis) characterized elbow flexion. C: X-axis is defined as being collinear to the upper arm axis; Z-axis is defined as being collinear to the radiocubital axis; Y-axis is defined as being orthogonal to the plane XZ; H (hand) is the center of gravity of the infrared emitting diodes 1 and 2. The angle 
between the X-axis and the projection of the wrist-hand vector in the XY plane defined wrist azimuth. The angle 
between the wrist-hand vector and the XY plane defined wrist elevation.
was determined on the other unperturbed curves, i.e., the point for which the Euclidean distance between the median (C20) and the other (C60, C40, C0, C20) curves was minimal. For each referred curve (C60, C40, C0, C20), Hotelling's T2 test (multivariate test for differences in means, Anderson 1958) was used to compare the three-dimensional positions (X, Y, Z) of P and P (Hotelling 1dynamic analysis). The same computational method was applied to compare the control (C20) and perturbed conditions (P60, P40, P0, P20) and to detect the earliest spatial point from which the perturbed trajectories began to deviate from the C20 control trajectory. The reaction time of the motor system to the perturbation (RTP) was estimated as the difference between the time of divergence minus the time of occurrence of the perturbation. To prevent erroneous detection of the divergence point between C20 and the other curves, a point P of C20 was considered as the divergence point only if the distance between the reference and referred curves remained significant for all the points of C20 following P.
, these "directional parameters" may be more sensitive than the wrist position in determining the reaction time of the sensory motor system to an external perturbation. To prevent noisy detection of the divergence point of the wrist velocity angles, the statistical analysis was performed only between the time to peak acceleration and the time to peak deceleration of the movement.
).
) and elevation () angles of the upper arm. These angles were computed in a bodily frame of reference centered at the shoulder (the X-axis was defined as the sagittal axis; the Y-axis was defined as the frontoparallel axis; the Z-axis was defined as orthogonal to X and Y).
). This angle was defined as being positive for internal rotations (0° < < 90°) and negative for external rotations (0° < < 90°).
, or prosupination angle; same sign convention as for the upper arm rotation).
). This angle was defined as being equal to 180° when the upper limb and the forearm were collinear.
) and elevation () angles of the wrist. These angles were computed in a bodily frame of reference centered at the wrist [X-axis was defined as the forearm axis; Z-axis was defined as the axis defined by the diodes 3 and 4 (radiocubital axis in Fig. 3A); Y-axis was defined as being orthogonal to X and Z].
static analysis).
) was computed from the angular position signal with the use of a least-square second-order polynomial method (window ±4 points). The same method was used to compute the angular acceleration from the velocity signal. The onset and the end of the movement were computed automatically from the acceleration signal with the use of the following threshold: hand acceleration = 5% of the maximal angular acceleration.
RESULTS
Abstract
Introduction
Methods
Results
Discussion
References
20°, 0°, 20°, 40°, and 60°) and session (2 levels: S1 blocked trials and S2 control trials) were far from statistical significance for all of the kinematic variables [F(4,20) < 1, P > 0.4]. Likewise, none of the kinematic parameters was significantly affected by the session factor [F(1,5) < 1.1, P > 0.3]. The same result was observed for the final arm posture. As shown by Hotelling tests, the arm configurations observed in S1 and S2 were not significantly different for a given final orientation of the object to grasp (P > 0.55).
20°, 0°, 20°, 40°, and 60°) and learning (2 levels: early trials and late trials) factors were far from statistical significance for all of the kinematic variables [F(4,20) < 0.55, P > 0.70]. Likewise, none of the kinematic parameters was significantly affected by the learning factor [F(1,5) < 0.50, P > 0.5]. In regard to the postural parameters, Hotelling tests showed that the final arm configurations were not significantly different for the early and late perturbed trials (P > 0.85).
20 unperturbed curves diverged significantly from the C20 reference (P < 0.05) were 230, 312, 389, and 264 ms, respectively (Hotelling 1). As shown in Fig. 4 (average values for the 6 subjects) and Fig. 5 (average value for a representative subject), the wrist path curvature depended strongly on the object's final orientation. This graphic observation was confirmed by statistical analysis showing that the wrist path curvature varied significantly with the object's final orientation. As demonstrated by an ANOVA performed on the unperturbed trials (ANOVA 1), the path curvature increased monotonically and significantly [F(4,20) = 13.68, P < 0.0001] when the object was tilted from the right to the left. Path curvature was maximal for C60 (0.20) intermediate for C20 (0.16), and minimal for C20 (0.11). The differences in wrist path curvatures were associated with significant differences in the wrists' final positions (Fig. 4). As demonstrated by Hotelling tests, the final position reached by the wrist (X, Y, Z) was significantly different for all the unperturbed conditions considered two by two (P < 0.02). The result appeared surprising at first glance, considering that the location of the center of mass of the object remained constant during the whole experiment. In fact, the significant differences in the final wrist locations were related to the participation of the proximal segments to the final hand orientation (when the subject held the object, the elevation of the upper arm induced a change in wrist position in the frontoparallel plane; see the postural analysis) and to the contribution of wrist flexion, which differed slightly according to object orientation (the wrist flexion induced a modification of the wrist position along the sagittal axis; see the postural analysis).
View larger version (87K):
[in a new window]
FIG. 4.
Three-dimensional representation of wrist and elbow trajectories in the task space (trajectories were averaged for the 6 subjects after temporal normalization; perspective view from behind and above). Three final object orientations are considered: 20° right (20R = 20°), 20° left (20L = 20°), and 60° left (60L = 60°). For the unperturbed trials (black lines), both wrist and elbow trajectories varied with the final orientation of the object. For the perturbed trials (white lines), the wrist trajectories smoothly diverged from the C20 control condition. Strikingly, perturbed trajectories terminated for both the wrist and the elbow at the endpoint of the corresponding unperturbed trajectories. Note that the scales are different for each axis to improve legibility.
View larger version (26K):
[in a new window]
FIG. 5.
Projections of mean wrist trajectories in both the horizontal (A and C) and sagittal (B and D) planes for a representative subject (subject 2; each curve was averaged after temporal synchronization of the individual trials; n = 10). A and B: unperturbed trials (5 conditions). C and D: perturbed trials and the C20 control condition; only the largest perturbed conditions (P60, P 20) are represented in the sagittal plane to preserve legibility. Open circles the C20 control curve: times of divergence of the unperturbed (C60, C40, C0, C20) and perturbed (P60, P40, P0, P20) referred curves for subject 2. Note for the unperturbed trials that wrist path curvatures differ significantly according to object orientation. In addition, note for the perturbed trials that wrist paths diverge smoothly from the C20 control trajectory.
20 curves diverged from the C20 reference condition (P < 0.05) were 355, 385, 410, and 350 ms, respectively (Hotelling 1). Interestingly, the error induced by the perturbation was almost fully corrected by the motor system. As shown by Hotelling tests, the final position reached by the wrist after the perturbation was not statistically different from the position reached during the unperturbed trial, presenting the same orientation as the one reached after perturbation (P > 0.60; see Fig. 4).
20, respectively (Hotelling 1). Because these RTPs values remained relatively high compared with the values reported in several double-step experiments (Gentillucci et al. 1992; Prablanc and Martin 1992; Soechting and Lacquaniti 1983), we tried to estimate the reaction time of the sensory motor system considering the spatial orientation of the velocity vector, which is known to constitute a very sensitive parameter to detect hand trajectory modifications (Prablanc and Martin 1992). As expected, the RTPs values computed from the instant of divergence between the angles of the tangential velocity vectors were consistently shorter (~45 ms) than the RTPs values obtained with respect to the spatiotemporal analysis. The values at which the velocity vectors related to the P60, P40, P0, and P20 curves diverged from the velocity vector related to the C20 reference curve (P < 0.05) were 288, 312, 336, and 320 ms, respectively.
View larger version (23K):
[in a new window]
FIG. 6.
Individual (- - -) and averaged ( ) velocity and acceleration profiles recorded for a representative subject (S2). Note the morphological similarity of the perturbed and unperturbed curves. In particular, note for the perturbed curves the absence of discontinuity or inflexion point that could be related to the modification of object orientation at movement onset. The observation of individual data shows that the smooth corrections observed during perturbed movements were not due to a computational artifact related to the averaging procedure.
View this table:
TABLE 1.
Postural parameters
View larger version (17K):
[in a new window]
FIG. 7.
Representation of all the final upper limb angles (means ± SD), according to object orientation, for a representative subject (subject 1). For unperturbed trials ( 
), a monotonic and nearly linear relationship can be observed between object tilt and joint angles. For the perturbed trials (
), the final value of the upper limb angles was very close to the final value obtained during the unperturbed trials, presenting the same orientation as those reached after perturbation. In other words, the posture reached by the arm was not different for a given orientation of the object to be grasped. Note that the vertical scales were adjusted for each panel to improve legibility.
View larger version (17K):
[in a new window]
FIG. 8.
Representation of the movement onset time (MO), movement end time (ME), and velocity-related kinematic landmarks (KL) as a function of the 7 upper limb angles (mean values, n = 6 subjects: , upper arm azimuth; , upper arm elevation; , upper arm rotation; , elbow flexion; , forearm rotation; , wrist azimuth; , wrist elevation). Velocity-related kinematic landmarks are the time to peack velocity for all the angles whose variations were monotonic during the course of the motion (, , , , , ). For the elbow joint (), which was successively flexed and extended during the movement, the time to movement reversal is represented instead of the time to peak velocity. For the sake of clarity, the time t = 0 is arbitrarily defined as 10 ms before the time of occurence of the earliest angular movement. With the exception of the wrist elevation, whose movement onset was slightly delayed with respect to the movement onset of the other joints, movement began nearly at the same time for all joint angles of the upper limb. This initial temporal synchronization held during the whole movement. Thus angular displacements ended almost concurrently on all the upper limb joints. Moreover, the time to peak velocity occurred nearly at the same time for all the angles whose variations were monotonic (as for movement onset, the temporal disparities were greater for the distal than for the proximal angles). Concerning the elbow joint, whose movement presented a reversal, the time of movement reversal was nearly coincident with the time to peak velocity of the other proximal angles.
View larger version (92K):
[in a new window]
View larger version (88K):
[in a new window]
FIG. 9.
Three-dimensional representation of movement trajectories in the joint space for a representative subject (subject 5; trajectories were averaged after temporal normalization: n = 10 individual trials). Only the extreme (C60: 
; P60: 
; C20: 
; P20: 
) and median (C20: ) conditions were represented to preserve legibility. Because of the existence of a temporal synchronization between the joint angles' individual variations, roughly straight paths were observed for the shoulder angles whose variations were monotonic (upper arm azimuth, upper arm elevation, upper arm rotation; A). Such was not the case for the elbow angle whose variations were not monotonic (B). When elbow angle was considered, consistently curved paths were observed in the joint space (note that the amount of path curvature varied as a function of the orientation of the object to be grasped). During the unperturbed trials, joint covariations patterns were reorganized. The perturbed trajectories diverged progressively from the C20 control trajectory to reach a new postural state that was nearly identical to the state reached when the object was initially presented along the orientation reached after perturbation.
20 and P60 conditions. As shown in this figure, the perturbed trajectories diverged progressively from the control trajectory because of a modification of the angular covariation ratios. This modification was tuned to bring the upper limb into the same angular configuration as the one reached during the corresponding unperturbed trial. Despite movement reorganization, the final temporal coupling observed among joint angles during the unperturbed trials was preserved during the perturbed trials. In other words, movements ended at approximately the same time on all joint angles during the perturbed movements (maximal range 25 ms for the P0 condition).
DISCUSSION
Abstract
Introduction
Methods
Results
Discussion
References
) or by segmenting the total movement into independent functional modules (Jeannerod 1988), but by linking the degrees of freedom of the upper limb in a fixed manner (Bernstein 1967). Our data suggest that the motor system solved the inverse mapping problem by implementing stereotyped postures for each orientation of the object. Thus, for the unperturbed trials, statistically identifiable upper limb configurations were observed for each orientation of the object to grasp. Moreover, for the perturbed trials, the final configuration of the upper limb was the same as that obtained when the object was initially presented along the final orientation reached after perturbation. These results are particularly remarkable considering that the final upper limb posture reached during perturbed and unperturbed trials could not be the result of mechanical constraints: as demonstrated by both empirical observations and the wide area of variations observed across experimental conditions for proximal and distal joints, a large set of comfortable intermediate postures (for instance between the 20° and 60° unperturbed trials postures for the 20° to 60° perturbed condition) was easily achievable for each given orientation of the object to grasp.
; Rosenbaum et al. 1995). This conclusion is in agreement with recent computational models (Rosenbaum et al. 1995), and with experimental findings from Soechting and colleagues (Helms-Tillery et al. 1991; Soechting and Flanders 1989a,b), who showed that the representation of a visual target in extrapersonal space had to be transformed into a kinesthetic representation of arm segment orientation before a goal-directed movement could be implemented. It is also corroborated by several recent studies showing that postural variables are taken into account by the motor system in planing the movement (Flanders et al. 1992; Helms-Tillery et al. 1991; Hore et al. 1992, 1994; Lacquaniti and Maioli 1994a,b; Miller et al. 1992; Rosenbaum et al. 1990, 1992; Scott and Kalaska 1995; Straumann et al. 1991).
; Berkinblit et al. 1986; Lacquaniti and Maioli 1994b; Prablanc and Martin 1992), which implies that a specific goal can be achieved with the use of different patterns of movement. The biological validity of this concept has been widely demonstrated since the initial works of Lashley (1930) and Bernstein (1967). For example, Abbs and Gracco (1984) observed that a given syllable could be correctly pronounced with the use of very different motor patterns. Likewise, Berkinblit et al. (1986) showed in the spinal frog that the wiping reflex, which consists of movements of the hindlimb directed toward a nociceptive stimulus, could generate different angular combinations for a final given position of the hindlimb endpoint.
, who were unable to reproduce the results of Berkinblit et al. (1986), reported the existence of a stereotyped pattern of angular covariations for wiping movements in the intact and spinal frog. In a similar vein, dynamic and static fixed angular patterns were identified in automatic responses to stance perturbations (Nashner and McCollum 1985), in perturbed and unperturbed pointing movements (Hore et al. 1992; Soechting and Lacquaniti 1981, 1983), in eye movements (Tweed and Vilis 1990), in throwing movements (Hore et al. 1994), and in eye-head-arm coordination (Straumann et al. 1991; Vercher et al. 1994). In fact, the absence of motor equivalence processes observed in the present experiment reinforces the idea that the final posture to be reached is the primary variable controlled by the nervous system during goal-directed movements performed in the absence of external constraints imposing a specific hand path.
). Regarding the prehension movements studied in the present experiment, two types of invariances can be expected: spatial or segmental. If the movement is planned in the task space (spatial invariance), i.e., if the subject tries to move the effector along a specific path (for instance a straight line path), strong variations should be observed in the joint covariations patterns. On the contrary, if the movement is planned in the joint space (segmental invariance), i.e., if the subject tries to preserve a specific pattern of joint covariations, strong variations should be observed in the external hand path. Our data agree with this second hypothesis. Indeed, the spatial path followed by the hand in the present experiment was both consistently curved and significantly variable with respect to the orientation of the object to be grasped. By contrast, the joint covariations patterns were very stable, irrespective of the experimental condition: during the unperturbed trials the individual variations of the upper limb angles were systematically coupled with respect to the time (joint synergies). Although effective for all the upper limb angles, the temporal synchronization was particularly consistent for the proximal angles (shoulder, elbow, and radiocubital joint angles). This latter point strongly suggests that the joint and spatial paths observed during the present experiment were not directly controlled by the motor system, but were the result of the synergic transition programmed between the starting and target postures. As initially underlined by Bernstein (1967) and Soechting and Lacquaniti (1981), the existence of such joint synergies is highly advantageous from a computational point of view. Indeed, a strict temporal coupling between anatomically independent degrees of freedom decreases drastically the movement control complexity by decreasing the number of independent parameters that have to be controlled by the system.
). In light of this remark, it appears that the movement reversal observed in the present experiment for the elbow joint resulted from the functional constraints imposed by the task. This result could be very interesting because it suggests that the postural control scheme can successfully deal with extrinsic requirements by implementing "angular via point," or more generally intermediate posture (Bizzi et al. 1992).
(110 ms); Soechting and Lacquaniti 1983 (110 ms) could be overestimated because of the smoothness of trajectory modifications. Indeed, the smoother the corrections, the more difficult their detection. This assumption is in agreement with the results reported by Prablanc and Martin (1992), who described longer RTPs for perturbations requiring an increase of movement curvature (reaction time ~ 274 ms) than for perturbations requiring a reversal of movement curvature (reaction time ~ 155 ms).
; Pélisson et al. 1986; Prablanc and Martin 1992).
; Gielen et al. 1984), the initial motor program has to be interrupted and replaced (Gentillucci et al. 1992; Massey et al. 1986; Paulignan et al. 1991). This assumption might account for the discrepancy existing between the smooth patterns of correction we observed in the present experiment and the jerky reorganizations reported in previous double-step studies also dealing with prehension movements (Carnahan et al. 1993; Gentilucci et al. 1992; Paulignan et al. 1991; Stelmach et al. 1994). As shown in those studies, which used larger perturbations than the ones applied in the present experiment, when prehension movements were perturbed at their onset the initial response toward the first target was interrupted very early (deceleration) before a second response toward the second target was initiated (reacceleration).
| |
ACKNOWLEDGEMENTS |
|---|
We thank Dr. M. Jordan for valuable comments on the manuscript. We also thank C. Urquizar and P. Monjaud for technical assistance. This work was supported by the Human Frontier Science Program.
| |
FOOTNOTES |
|---|
Address for reprint requests: C. Prablanc, INSERM Unité 94, 69500 Bron, France.
Received 15 March 1996; accepted in final form 25 September 1996.
| |
REFERENCES |
|---|
|
Abstract
Introduction
Methods
Results
Discussion
References
|
|---|
A Silver Jubilee: Synergies in Experimental Psychology, Artificial Intelligence and Cognitive Neuroscience,
edited by
D. E. Meyer,
and S. Kornblum
. Cambridge, MA: MIT Press, 1992, p. 803-820This article has been cited by other articles:
|
|
 |
|
  N. Gosselin-Kessiby, J. Messier, and J. F. Kalaska Evidence for Automatic On-Line Adjustments of Hand Orientation During Natural Reaching Movements to Stationary Targets J Neurophysiol, April 1, 2008; 99(4): 1653 - 1671. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Malfait, D. Y. Henriques, and P. L. Gribble Shape Distortion Produced by Isolated Mismatch Between Vision and Proprioception J Neurophysiol, January 1, 2008; 99(1): 231 - 243. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. N. Aflalo and M. S. A. Graziano Relationship between Unconstrained Arm Movements and Single-Neuron Firing in the Macaque Motor Cortex J. Neurosci., March 14, 2007; 27(11): 2760 - 2780. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. M. Beurze, S. Van Pelt, and W. P. Medendorp Behavioral Reference Frames for Planning Human Reaching Movements J Neurophysiol, July 1, 2006; 96(1): 352 - 362. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. N. Aflalo and M. S. A. Graziano Partial tuning of motor cortex neurons to final posture in a free-moving paradigm PNAS, February 21, 2006; 103(8): 2909 - 2914. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Vindras, M. Desmurget, and P. Viviani Error Parsing in Visuomotor Pointing Reveals Independent Processing of Amplitude and Direction J Neurophysiol, August 1, 2005; 94(2): 1212 - 1224. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Desmurget, V. Gaveau, P. Vindras, R. S. Turner, E. Broussolle, and S. Thobois On-line motor control in patients with Parkinson's disease Brain, August 1, 2004; 127(8): 1755 - 1773. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. B. Torres and D. Zipser Reaching to Grasp With a Multi-Jointed Arm. I. Computational Model J Neurophysiol, November 1, 2002; 88(5): 2355 - 2367. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Desmurget, H. Grea, J. S. Grethe, C. Prablanc, G. E. Alexander, and S. T. Grafton Functional Anatomy of Nonvisual Feedback Loops during Reaching: A Positron Emission Tomography Study J. Neurosci., April 15, 2001; 21(8): 2919 - 2928. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
