Loading field map
A stroke survivor trying to lift a cup. A child with cerebral palsy learning to step off a curb. A veteran with a spinal cord injury practicing a transfer from bed to chair. These are the moments that define motor learning in rehabilitation. The central question is deceptively simple: what changes inside a person when they practice a movement, and how can a therapist best guide that change? For nearly a century, rehabilitation researchers have offered competing answers, and the debate between them has reshaped everything from the design of therapy sessions to the equipment in a clinic gym.
For the first several decades of modern rehabilitation, the dominant answer came from a family of techniques now called Neurofacilitation Approaches. These methods—including the Bobath concept, Brunnstrom's movement therapy, and proprioceptive neuromuscular facilitation—shared a core assumption: the nervous system had been damaged, and the therapist's job was to manually guide the correct movement pattern so that the brain could relearn it. The therapist positioned the limb, stimulated key reflexes, and inhibited abnormal tone, all based on a hierarchical model of motor control in which higher centers normally suppress lower, reflexive ones. Recovery meant restoring that hierarchy.
Neurofacilitation Approaches aligned closely with the Medical Model of Disability, which treated impairment as a problem located inside the individual body, to be fixed or compensated for by expert intervention. The unit of analysis was the reflex arc or the muscle synergy, and the evidence that practitioners valued came from clinical observation and case reports. The learner was largely passive, receiving input from the therapist. By the 1970s, however, a growing body of research began to challenge this picture. Studies showed that the specific techniques of neurofacilitation did not consistently produce better outcomes than simpler, more active forms of practice. The field needed a new way to think about what learning actually required.
Task-Oriented Motor Learning emerged directly from this pressure. Drawing on cognitive psychology and the schema theory of motor control developed by Richard Schmidt, this framework argued that motor learning is not about training reflexes but about acquiring generalizable movement strategies. A person learning to reach after a stroke does not need to have their arm passively moved through the correct arc; they need to practice the task of reaching itself, in varied contexts, so that they build a flexible internal model—a schema—that can adapt to new cups, new distances, and new postures.
This was a fundamental shift. The therapist became a coach and a task designer rather than a manual guide. The learner was an active problem-solver, and the unit of analysis was the functional task, not the muscle synergy. Evidence came from controlled experiments measuring retention and transfer of skills, and the framework found strong support in studies showing that task-specific practice—practicing the whole movement in a realistic context—produced better long-term outcomes than repetitive, decontextualized exercises. Task-Oriented Motor Learning did not fully replace Neurofacilitation Approaches; many clinicians continued to use facilitation techniques for specific goals, especially in early recovery. But the cognitive framework redefined what counted as learning and who was responsible for it.
Just as Task-Oriented Motor Learning was gaining traction, a more radical challenge arrived from an unexpected direction. Dynamical Systems Theory, rooted in nonlinear physics and the work of Esther Thelen and others on infant motor development, proposed that motor behavior does not require internal schemas or representations at all. Instead, movement patterns emerge spontaneously from the interaction of multiple subsystems—the body's biomechanics, the environment's affordances, the task's demands—without a central executive giving commands. A toddler learning to walk does not have a stored walking program; walking self-organizes as the child's strength, balance, and the surface underfoot interact.
For rehabilitation, this was a deeply unsettling idea. If movement is self-organized, then the therapist's role is not to teach a correct pattern or even to train a cognitive schema, but to manipulate the constraints that shape emergence: change the surface, adjust the load, alter the visual feedback, and let the learner's system find its own solution. Dynamical Systems Theory challenged the very necessity of cognitive schemas, arguing that what looks like a stored plan is actually a stable attractor state in a dynamical system. The unit of analysis shifted to the whole person-environment system, and evidence came from kinematic analyses of movement variability and stability over time.
Task-Oriented Motor Learning and Dynamical Systems Theory both remain active and influential in rehabilitation research and practice today. They agree on several key points that distinguish them from the older Neurofacilitation Approaches: practice must be active, task-specific, and varied; the learner's environment matters; and the therapist should facilitate rather than control. Both frameworks have contributed to the rise of intensive, task-based interventions such as constraint-induced movement therapy and locomotor training.
Where they disagree is more fundamental. Task-Oriented Motor Learning retains a cognitive architecture: it assumes that the learner forms, stores, and retrieves internal representations of movement. This makes it compatible with explicit instruction—telling a patient "reach farther" or "twist your wrist"—and with structured feedback schedules. Dynamical Systems Theory, by contrast, treats explicit instruction as just another constraint on the system, not a privileged source of learning. It predicts that implicit learning, in which the learner discovers a movement solution without being told the rule, is often more robust and transferable. The two frameworks also differ on the role of variability: Task-Oriented Motor Learning sees variability as noise to be reduced through practice, while Dynamical Systems Theory sees it as the raw material from which new, adaptive patterns emerge.
No single framework has won the argument. In contemporary rehabilitation science, researchers and clinicians often draw on both perspectives, depending on the population and the goal. For a patient with a recent stroke who needs to regain basic reaching, a task-oriented approach with explicit goals and feedback may be most efficient. For a child with developmental coordination disorder who struggles to adapt to changing playground equipment, a dynamical systems approach that varies constraints and lets the child explore may be more effective. Some research groups are actively working on integrated models that treat cognitive representations as emergent properties of dynamical systems, though this synthesis remains controversial.
The history of motor learning in rehabilitation is not a story of one framework replacing another. It is a story of an evolving argument about what movement is and how it changes. Neurofacilitation Approaches gave the field its first systematic methods and a focus on the nervous system. Task-Oriented Motor Learning gave it a cognitive, learner-centered model and a rigorous experimental evidence base. Dynamical Systems Theory gave it a radical alternative that questions whether the brain stores movement at all. Together, they define the questions that drive the field today: How much of motor learning is inside the head, and how much is in the world? And what does a therapist actually do when the system learns itself?