Secondary Logo

Journal Logo


Handedness: Differential Specializations for Control of Trajectory and Position

Sainburg, Robert L.

Author Information
Exercise and Sport Sciences Reviews: October 2005 - Volume 33 - Issue 4 - p 206-213
  • Free


Cerebral Lateralization

Lateral specialization is a ubiquitous feature of cerebral organization in humans. Research by Gazzaniga (4) on patients with “disconnection syndrome” has shown that the right and left hemispheres have strong advantages for different functions. For example, semantic features of language are specialized within the left hemisphere, whereas certain visual–spatial analyses are specialized within the right hemisphere. Such coopting of specific processes by one or the other hemisphere is thought to expand function by reducing redundancy, thereby increasing available neurons for local circuits. This results in a “no cost extension” by which the system can increase function without requiring more neural tissue (4).


Handedness might be the most prominent example of lateral specialization, occurring in approximately 96% of the human population (1). Although the neural foundations of handedness remain incompletely understood, it has been well established that handedness is reflected by substantial asymmetries in neural structure and functional activation (10). Some theories suggest that handedness reflects a nonfunctional artifact of a primary asymmetry in systems that govern language, cognition, and/or emotion (1). An alternative explanation, proposed here, is that handedness confers direct advantages to the control of movements, reflecting an organizing principle for movement control.

Dynamic Dominance Hypothesis

Many attempts have been made to distinguish dominant from nondominant arm control in terms of open-loop or closed-loop processes. This research was inspired by the distinction elaborated by Woodworth (15), who described the acceleration phase of rapid aimed movements as reflecting preplanned control (open-loop processes), and the deceleration phase as largely influenced by sensory feedback (closed-loop processes). However, this body of literature has generally failed to establish differences in performance that can be reliably attributed to such a division of labor (10). Handedness continues to be described as a unidimensional trait, reflected simply by the tendency to prefer the dominant limb for certain tasks. Handedness is currently identified by inventories (1,7,10) that determine the tendency to use one or the other arm in performing a given set of tasks, such as writing, throwing a ball, and striking a match.

Recent research from our laboratory provides evidence that handedness might result from a distinct, but complementary division of resources: dominant system specialization for controlling limb trajectory, and nondominant specialization for control of limb position. We term this hypothesis dynamic dominance, because of evidence that dominant arm control entails more efficient and accurate coordination of muscle actions with the forces that arise between the moving segments of the multijoint limb, or interaction forces. Recent findings indicate that the nondominant limb system shows advantages for positional control, even in the face of unexpected loads applied during the course of movement. In this article, studies from our laboratory that differentiatedominant and nondominant systems in terms of specialization of trajectory and positional control mechanisms during reaching movements are reviewed.


The experimental set-up for these studies is depicted in Figure 1. Subjects sit facing a table with their arm supported over the horizontal surface by an air-jet system. A back-projection screen is positioned above a mirror, which reflects a two-dimensional virtual reality environment, in which a start position and target are presented. Subjects are instructed to move the hand to a displayed target in response to a “go” signal. Feedback of the fingertip position is given while placing the hand at the start position, but can be removed at the “go” signal. Positions and orientations of each segment are sampled using a Flock of Birds® (FOB; Ascension Technology) magnetic 6-df movement recording system, and are synchronized with electromyographic (EMG) recordings.

Figure 1.
Figure 1.:
Experimental set-up, side view and top view (see text). Flock of Birds (FOB: Ascension Technologies) sensors.


Lateralization of Intersegmental Coordination

To coordinate the multiple segments of the limb, muscle actions must be produced to account for intersegmental interaction torques, the dynamic interactions arising between the segments of the moving limb. We designed a reaching task that would elicit progressively greater interaction torques at the elbow joint by varying the required shoulder motion while keeping the required elbow motion constant (11,2). Figure 2 shows three targets with representative hand paths. Final position accuracies were slightly more accurate for the nondominant arm. However, the hand paths and respective joint coordination patterns were systematically different. Dominant hand paths showed gentle medial to lateral curvatures for all target directions, whereas nondominant hand paths showed oppositely directed curvatures that increased in magnitude across directions.

Figure 2.
Figure 2.:
A. Sample dominant (black) and nondominant (gray) hand paths plotted in right-hand coordinates. B. Elbow joint torques for the trials displayed as dotted lines to target 1 above. [Adapted from Sainburg, R.L., and D. Kalakanis. Differences in control of limb dynamics during dominant and non-dominant arm reaching. J. Neurophysiol. 83:2661–2675, 2000. Copyright © 2000 The American Physiological Society. Used with permission.]

Analysis of limb segment torques revealed substantial differences in coordination. Figure 2B shows the dominant and nondominant elbow torques that correspond to the two dashed hand paths shown for target 1. The net joint torque is the sum of two components: interaction torque, resulting from motion of the attached segment, and muscle torque, largely representing the effects of the muscles crossing a joint (2). It must be stressed that this calculation is based on a rigid body model of the limb. Neither the contributions of noncontractile tissues nor the effects of muscle coactivations can be differentiated using this analysis alone. The net torque varies with joint acceleration and reflects the action of muscle and interaction torques in accelerating the segment about the joint in question.

Note that the nondominant hand path to target 1 is straight, whereas the dominant path is gently curved medial to lateral (dashed paths; Fig. 2A). The curvature of the dominant path was associated with forward flexion at the shoulder, which produced substantial elbow interaction torque. Thus, elbow muscle torque contributed only approximately 50% to elbow net torque, whereas interaction torque contributed the other 50%. In contrast, the straight path of the nondominant arm was associated with little shoulder motion, and thus a very small elbow interaction torque. As such, nondominant elbow acceleration was driven almost entirely by elbow muscle torque, reflecting a less torque-efficient control strategy.

Across targets, the dominant and nondominant systems produced movements with similar speeds and accuracies, but dominant arm motions were made with a fraction of the muscle torque as those of the nondominant arm. In fact, in movements that were matched for speed and displacement, the dominant arm often used less than half the muscle torque measured in the nondominant arm (11,2). This emphasizes the fact that the coordination differences between the limbs are not simply caused by differences in torque generating capacities. In our tasks, nondominant arm movements consistently showed larger muscle torque contributions to net torque and, thus, less efficient movements. These results have been corroborated by electromyographic (EMG) recordings in multijoint movements, which revealed corresponding differences in normalized EMG activities between the limbs (2).

These data suggest that the dominant system can better anticipate, and thus take advantage of, intersegmental dynamics, when planning movements. If so, one might expect hand path curvatures to be independent of the amplitude of these interactions. Figure 3 shows a measure of hand path curvature, the average direction change between the acceleration and deceleration phases of motion, regressed across elbow joint interaction torque impulse. As expected, dominant hand path curvatures were independent of the magnitude of interaction torques. This is reflected in Figure 2 by the similar dominant arm curvatures across the three targets and the flat relationship shown in the scatter-plot (Fig. 3, right). In contrast, nondominant hand path curvatures were effectively enslaved to these interactions, as reflected by the increasing curvatures across targets in Figure 2, and the steep slope and strong relationship of Figure 3, left. These findings support the idea that the dominant hemisphere/limb system is specialized for anticipating and controlling limb dynamics.

Figure 3.
Figure 3.:
Dependence of hand path direction change, a measure of curvature, and interaction torque impulse for nondominant (left) and dominant (right) arms. [Adapted from from Sainburg, R.L., and D. Kalakanis. Differences in control of limb dynamics during dominant and non-dominant arm reaching. J. Neurophysiol. 83:2661–2675, 2000. Copyright © 2000 The American Physiological Society. Used with permission.]

Dominant Limb Advantages are Specific to Controlling Dynamics

Because the previous findings emphasize a dominant system specialization for control of limb dynamics, we hypothesized that features of control that do not stress intersegmental dynamics should not elicit dominant arm advantages. In a direct test of this prediction (10), we compared adaptation to an eccentrically positioned inertial load, which produced novel interaction torques, with that of adaptation to a visuomotor rotation, which did not alter task dynamics. Subjects performed a center-out reaching task in eight directions, adapting either to a 30° visuomotor rotation, or an eccentrically positioned inertial load, attached to the forearm brace. Both manipulations resulted in large curvatures in the movements, which were progressively reduced through the course of adaptation. Figure 4A shows example hand paths from the first (solid) and last (dotted) cycles of the adaptation session. Figure 4B shows our measure of linearity (minor/major axis), plotted against movement cycle (mean of eight trials) across the adaptation session, and averaged across all subjects. For both adaptation conditions, the error progressively decreased as subjects adapted to the task. However, as predicted by our hypothesis, the dominant arm showed more complete adaptation to the inertial load, whereas both arms showed similar visuomotor adaptations. Thus, the advantages of the dominant system seem specific to the control of limb dynamics.

Figure 4.
Figure 4.:
A. Sample hand paths for the first trial (dotted) and a late trial (solid) from the adaptation session, in which subjects are exposed to either the inertial load (left) or the visuomotor rotation (right). B. Measure of linearity for each cycle (eight trials), averaged across all subjects during the course of the session. The shaded cycles reflect those from which the sample trials (top) were taken. [Adapted from Sainburg, R.L. Evidence for a dynamic-dominance hypothesis of handedness. Exp. Brain Res. 142:241–258, 2002. Copyright © 2002 Springer-Verlag GmbH. Used with permission.]

Nondominant Specialization for Control of Limb Position

Because few functional advantages in nondominant limb performance have previously been identified, the nondominant system has traditionally been viewed as a naïve, unpracticed, analog of the dominant hemisphere/limb system. However, regardless of the less efficient trajectories of the nondominant arm, final positions are often similar or even more accurate than that of the dominant arm, a consistent finding across many of our studies (2,3,10,11). This suggests that the dominant system may compromise final position accuracy in improving trajectory efficiency, whereas the nondominant system may be more concerned with final position control.

In a direct test of this hypothesis, we assessed interlimb asymmetries in responses to an unexpected inertial load applied during targeted single-joint movements (3). Subjects made repetitive 20° elbow flexion movements, using a modification of the experimental set-up depicted in Figure 1. For this study, the upper arm was immobilized by a brace attached to the table. On random trials, a 2-kg mass was attached to the forearm splint before the “go” signal. No information about the mass was available before movement, and thus EMG activity recorded before peak tangential hand acceleration was the same for loaded and baseline trials. After this point, substantial changes in muscle activity occurred.

Figure 5A shows ensemble averaged displacement and velocity profiles from baseline and loaded trials, whereas Figure 5B shows the corresponding EMG profiles. The effect of the load in reducing peak velocity was the same for both arms, but the response of each limb to this perturbation was different. For both arms, the load compensation response was associated with a reduction in extensor muscle activity (triceps), resulting in a prolonged flexor velocity phase of motion (time between the arrows in Figure 5A inset). The extensor response is reflected in the slightly smaller anconeus burst for loaded trials in Figure 5B. For the nondominant arm, this resulted in effective load compensation, such that no significant differences in final position accuracy occurred between loaded and baseline trials. However, the dominant arm response also included a considerable increase in flexor muscle activity (circled in Fig. 5B), which substantially prolonged the flexor acceleration phase of motion. As a result, the dominant arm overcompensated for the effects of the load, producing a large and systematic overshoot of final position. Whereas the dominant arm shows more efficient coordination for multijoint movements, this was not the case for the feedback-mediated responses to mechanical loads in the single joint movements studied here. We suggest that the dominant arm is specialized for feed-forward control of trajectory features, such as intersegmental dynamics, whereas the nondominant arm is specialized for feedback mediated control of position.

Figure 5.
Figure 5.:
A. Representative displacement and velocity (insets) for load compensation paradigm. Data have been normalized to baseline performance (solid lines). All data have been synchronized to peak tangential hand acceleration and averaged. B. Corresponding EMG data from biceps brachii (positive) and anconeus (negative). The x-axis for EMG data is expanded, relative to the kinematic data. [Adapted from Bagesteiro, L.B., and R.L. Sainburg. Non-dominant advantages in load compensation during rapid elbow joint movements. J. Neurophysiol. 90:1503–1513, 2003. Copyright © 2003 The American Physiological Society. Used with permission.]

These findings support the hypothesis that the nondominant system is better adapted for controlling limb position. The responses to the load began approximately 50 ms after movement onset, consistent with the latency of segmental reflexes. We thus expect interlimb differences in this study to result from differences in the gain modulation of such feedback circuits. It is plausible that the nondominant system may adjust such gains in accord with positional signals, whereas the dominant limb may be more responsive to velocity-dependent signals. Differentiation of position and velocity information could be mediated by differential regulation of static and dynamic gamma motor neurons, as elaborated by Prochazka (9). This might account for the similar latency, yet higher gain responses of the dominant arm, as well as the lower efficacy of dominant arm responses in achieving an accurate final position.

Interlimb Differences in Control of Movement Extent

Our findings on multijoint movements did not distinguish whether dominant arm advantages were restricted to coordination of intersegmental dynamics or reflected a more general specialization for trajectory control mechanisms. We thus examined interlimb differences in targeting movement extent in single joint movements at the elbow (12). For such movements, peak velocity shows a strong correlation with movement distance, indicating that distance is specified early in the movement. Two aspects of the acceleration profile, the amplitude and duration, have been associated with distinct control mechanisms, characterized by the pulse-step model as pulse-height and pulse-width, respectively. Although pulse-height occurs too early to be affected by sensory feedback, pulse-width has been shown to be extensively modulated by feedback, and has been hypothesized to correct for inaccuracies in pulse-height control in accord with the intended final position (5,12). We hypothesize that dominant arm performance might be better-adapted for pulse-height control, whereas nondominant performance might be better adapted for pulse-width control.

We tested this hypothesis using a single joint reaching task, requiring rapid elbow extension movements toward a range of target distances (12). Figure 6 shows the velocities and accelerations for a series of progressively longer movements. Although both arms scaled peak movement speed with movement distance, the two limbs achieved this scaling through qualitatively different mechanisms.

Figure 6.
Figure 6.:
Ensemble averaged tangential hand velocity (A) and acceleration (B) profiles. All profiles have been synchronized to peak tangential hand acceleration (zero on x-axis). [Adapted from Sainburg, R.L., and S.Y. Schaefer. Interlimb differences in control of movement extent. J. Neurophysiol. 92:1374–1383, 2004. Copyright © 2004 The American Physiological Society. Used with permission.]

As can be seen by the acceleration plots of Figure 6, the initial amplitude of joint acceleration scaled with peak speed in the dominant arm only. This reflects a predictive mechanism referred to as “pulse-height control” (5). In contrast, the nondominant arm initiated each movement with a stereotypical acceleration amplitude, regardless of intended movement distance. Instead of varying acceleration with movement speed, the nondominant system varied the duration of acceleration in order to accomplish the scaling of velocity with distance. This “pulse-width” mechanism has been attributed to somatosensory based feedback mechanisms (5,12). In this case, such feedback apparently adjusted the duration of joint torque in accord with the intended final position. This strategy likely reflects online control processes that are responsive to imposed forces, such as are required when stabilizing objects against loads imposed by dominant arm actions.

Interlimb Transfer of Motor Learning

The idea that each hemisphere/limb system is specialized for independent features of task performance leads to the question of how the two systems share information with one another. Previous research on interlimb transfer of motor learning has indicated that transfer is usually asymmetrical and that the direction of greatest transfer can vary with the task (13). We asked whether this asymmetry would occur for distinct features of performance within a given task by examining adaptation to a novel visuomotor rotation (13).

Two groups, comprising seven right-handed subjects each, adapted to a 30° counterclockwise rotation in the visual display during a center-out reaching task performed to eight directions. Each group first adapted using either the right (RL) or left (LR) arm, followed by opposite arm adaptation. To assess transfer, we compared the same side arm movements (either right or left) after opposite arm adaptation with those performed before opposite arm adaptation. Figure 7 shows measures of initial movement direction and final position error, averaged across all subjects, during the course of adaptation. As expected, different features of movement transferred in different directions: opposite arm training improved only the initial direction of dominant arm movements, whereas it improved only the final position of nondominant arm movements.

Figure 7.
Figure 7.:
Averaged direction error (top) and final position error (bottom) for dominant and nondominant arms during the course of the adaptation session for both LR (solid) and RL (open) groups. For the dominant arm, LR data shows the effect of opposite arm training, whereas for the nondominant arm RL data shows the effect of opposite arm training. [Adapted from Sainburg, R.L., and J. Wang. Interlimb transfer of visuomotor rotations: independence of direction and final position information. Exp. Brain Res. 145:437–447, 2003. Copyright © 2003 Springer-Verlag GmbH. Used with permission.]

These results support the hypothesis that each arm controller has access to information learned during opposite arm training. However, each controller appears to use this information differently, depending on its proficiency for specifying particular features of movement. The dominant arm can use information gained during nondominant training to plan the direction of movement in accord with the visuomotor rotation. However, this improvement does not impart better final position accuracy, suggesting that such direction planning is independent of the mechanisms responsible for correcting to the final position. In contrast, opposite arm adaptation improves nondominant arm final position accuracy, but not initial direction accuracy. Each hemisphere/limb system may compromise other features of control in improving the aspect of control for which it is best adapted.


The findings described here indicate that each hemisphere/limb system is specialized in controlling different features of movement. This is achieved through distinct neural mechanisms. The dominant system for controlling limb trajectory and the nondominant system appears specialized for controlling limb position. This division of labor is consistent with typical patterns of arm use during bimanual tasks, such as cutting bread or hammering nails, when the nondominant arm tends to stabilize an object against loads imposed by the dominant arm. In fact, Healey, Liederman, and Geschwind (7) previously described similar distinctions in dominant and nondominant limb use for a wide range of tasks across a large number of subjects. Rather than view the dominant system as generally “better” than its nondominant analog, we suggest that each system is specialized for unique processes (Fig. 8). This hypothesis is supported by findings in unilaterally lesioned stroke patients, which have revealed consistent deficits in the ipsilesional arm. Dominant hemisphere lesions produce deficits in trajectory speed, whereas nondominant lesions produce deficits in final positional accuracy (6).

Figure 8.
Figure 8.:
Schematic showing differentiation of dominant and nondominant hemisphere/limb systems for different features of control. Whereas both systems might contain trajectory and positional controllers, the positional controller is better-developed in the nondominant hemisphere, and the trajectory controller is better developed in the dominant hemisphere. The trajectory controller relies most heavily on feed-forward mechanisms that specify torque amplitude in accord with anticipated task dynamics. The position controller modulates torque duration through largely feedback mechanisms. However, some feed-forward contributions to control of torque duration have also been demonstrated.

The relative advantages of each hemisphere/limb system for different, but complementary functions suggest that lateralization allows greater adaptation through expanding the neural resources available for each process. It is thus likely that handedness developed in response to a need for more precise coordination, rather than as an artifact of lateralization in other systems, such as language. This is consistent with the idea that handedness may have emerged, phylogenetically, before the development of language functions, and also that it is expressed in animals who have not developed language (8). One important implication of this hypothesis is that failure to develop strong hand preference might reflect a failure to optimize the control system, and thus might be associated with limitations in control. In particular, subjects who show weak hand preference might be at a disadvantage for performance of tasks that stress coordination of multisegment dynamics with the dominant limb, as well as in tasks that stress positional control with the nondominant limb. This may account for the finding that a number of developmental coordination disorders have been associated with a failure to establish consistent hand preference (14). Further research is necessary to examine the effect of motor lateralization across different populations of lateral expression, including left-handers and mixed handers.


1. Annett, M. Handedness and cerebral dominance: the right shift theory. J. Neuropsych. Clin. Neurosci. 10:459–469, 1988.
2. Bagesteiro, L.B., and R.L. Sainburg. Handedness: dominant arm advantages in control of limb dynamics. J. Neurophysiol. 88:2408–2421, 2002.
3. Bagesteiro, L.B., and R.L. Sainburg. Non-dominant advantages in load compensation during rapid elbow joint movements. J. Neurophysiol. 90:1503–1513, 2003.
4. Gazzaniga, M.S. Cerebral specialization and interhemispheric communication: does the corpus callosum enable the human condition? Brain 123(pt 7):1293–1326, 2000.
5. Ghez, C., W. Hening, and J. Gordon. Organization of voluntary movement. Curr. Opin. Neurobiol. 1:664–671, 1991.
6. Haaland, K.Y., J.L. Prestopnik, R.T. Knight, and R.R. Lee. Hemispheric asymmetries for kinematic and positional aspects of reaching. Brain 127:1145–1158, 2004.
7. Healey, J.M., J. Leiderman, and N. Geschwind. Handedness is not a unidiminesional trait. Cortex 22:33–53, 1986.
8. Hopkins, W.D., and C. Cantalupo. Handedness in chimpanzees (Pan roglodytes) is associated with asymmetries of the primary motor cortex but not with homolgous language areas. Behav. Neurosci. 118:1176–1183, 2004.
9. Prochazka, A., M. Hulliger, P. Trend, and N. Dürmüller. Dynamic and static fusimotor set in various behavioural contexts. In: Mechanoreceptors: Development Structure and Function, edited by P. Hnik, T. Soukup, R. Vejsada, and J. Zelena. New York: Plenum Press, 1988, p. 417–430.
10. Sainburg, R.L. Evidence for a dynamic-dominance hypothesis of handedness. Exp. Brain Res. 142:241–258, 2002.
11. Sainburg, R.L., and D. Kalakanis. Differences in control of limb dynamics during dominant and non-dominant arm reaching. J. Neurophysiol. 83:2661–2675, 2000.
12. Sainburg, R.L., and S.Y. Schaefer. Interlimb differences in control of movement extent. J. Neurophysiol. 92:1374–1383, 2004.
13. Sainburg, R.L., and J. Wang. Interlimb transfer of visuomotor rotations: independence of direction and final position information. Exp. Brain Res. 145:437–447, 2002.
14. Signmundsson, H., and H.T. Whiting. Hand preference in children with developmental coordination disorders: cause and effect? Brain Cogn. 49:45–53, 2002.
15. Woodworth, R.S. The accuracy of voluntary movement. Psychol. Rev. 3:1–114, 1899.

handedness; lateralization; cerebral specialization; limb dynamics; limb dominance

©2005 The American College of Sports Medicine