ReviewThe planning and control of reaching movements
Introduction
The study of volitional movement has undergone rapid progress over the past several years. A major factor in this success has been the application of new theoretical ideas to the design and analysis of psychophysical investigations. Chief among these is the notion of internal models, hypothetical computations in the brain that either predict the outcome of some motor event (forward models) or calculate the motor command required to achieve some desired state (inverse models or feedforward controllers) [1]. Another key factor is the increased availability of functional imaging and transcranial magnetic stimulation (TMS), techniques that allow researchers to identify the neural substrates underlying complex behavioral phenomena. As a result, it has become easier to interpret psychophysical and computational findings in terms of our growing understanding of the neurophysiology of the sensorimotor pathways [2].
In the past year, significant achievements have been made in characterizing specific feedforward and feedback control structures involved in reaching. In particular, new results have clarified our understanding of the role of visual feedback in the early stages of reach planning and the ability to precisely control the complex dynamics of multijoint movements. Other major themes have been the role of learning in the maintenance of internal models and the manner in which intrinsic (e.g. joint, muscle) information and extrinsic (e.g. perceptual, task-specific) information are combined to form a motor plan. In this review, I will discuss current trends in the study of goal-directed reaching, focusing on these recent results.
Section snippets
Feedback control: on-line trajectory control in the parietal cortex
Visually guided reaching begins with the selection of a target from the visual scene and the formation of a movement plan. Recent studies of the posterior parietal cortex (PPC) and superior parietal lobe (SPL) in monkey have demonstrated that these areas contain the required combination of visual, somatosensory, and motor signals to be able to coordinate this first step [3], [4], [5], [6], [7]. New findings show that these early planning areas are not only responsible for the initial target
Feedforward control: interaction torques and internal models
Whereas the early stages of visually guided movement are seen to be under the control of feedback loops, evidence is mounting that aspects of the later stages of motor control rely less on feedback than many researchers previously thought. Precise control of multijoint movements requires control of the interaction torques that arise when the motion of one joint causes acceleration at another [38]. The seminal work of Sainburg and Ghez [39], [40] showed that patients who lack proprioception were
Learning and multiple internal models
In the examples of the previous section, the inverse model is adapted in the face of motor error. It is plausible, however, that other components of the system could also adapt. A possible example comes from a case in which visual feedback plays an important role in computing the inverse dynamics of the arm. When reaching while rotating the torso, the arm is subjected to Coriolis forces, yet movement of the torso does not disrupt accurate reaching. This situation is formally similar to the
Task-dependent optimal control
The task of the feedforward controllers discussed above is to allow for accurate control of the arm, which implies the existence of a motor plan. Recent thinking on the planning of reaching movements has been deeply influenced by the ‘minimum variance’ model of Harris and Wolpert [65], [66. They posit a signal-dependent variability in movement control, in which variability scales with the magnitude of the command signal. Movement trajectories are then chosen to minimize the resulting end-point
Conclusions
Over the past year, researchers studying goal-directed reaching have made significant progress, capitalizing on recently developed theoretical and experimental tools. We have a better understanding of high-level visual feedback loops and of low-level feedforward mechanisms for controlling multijoint movements. The notion of internal models has been central in these advances. In most of this work, evidence of predictive control or adaptive state estimation is used to identify internal models.
Acknowledgements
The author is grateful to Daniel Wolpert for helpful comments on a draft of this manuscript.
References and recommended reading
Papers of particular interest, published within the annual period of review,have been highlighted as:
- •
of special interest
- •
of outstanding interest
References (75)
Internal models for motor control and trajectory planning
Curr Opin Neurobiol
(1999)- et al.
Cortical control of reaching movements
Curr Opin Neurobiol
(1997) - et al.
Parietal cortex: from sight to action
Curr Opin Neurobiol
(1997) - et al.
Forward models — supervised learning with a distal teacher
Cog Sci
(1992) - et al.
Multiple paired forward and inverse models for motor control
Neural Networks
(1998) Inputs from motor and premotor cortex to the superior colliculus of the macaque monkey
Behav Brain Res
(1985)- et al.
Deficits in phasic muscle force generation explain insufficient compensation for interaction torque in cerebellar patients
Neurosci Lett
(1999) - et al.
Coding of intention in the posterior parietal cortex
Nature
(1997) - et al.
Reach plans in eye-centered coordinates
Science
(1999) - et al.
Early coding of reaching: frontal and parietal association connections of parieto-occipital cortex
Eur J Neurosci
(1999)