Author Response to Macpherson et al

We thank Macpherson et al¹ for their interest in our manuscript and the editors for the opportunity to respond. We hope that the continued conversation about motor learning in physical therapy will lead to a comprehensive understanding of what is currently known about this important topic and, ultimately, produce improved patient outcomes. In this response, we aim to provide clarity about the different mechanisms of motor learning and how they can be used to guide physical therapist practice and education.

First, we agree with Macpherson et al that these 4 motor learning mechanisms (use-dependent, instructive, reinforcement, and sensorimotor adaptation-based motor learning) likely co-occur to drive changes in movement, and the relative weight of each can change depending on the task constraints or stage of practice. However, current evidence suggests that these mechanisms are distinct from one another (for examples, see Roemmich et al² and Diedrichsen et al³) and engage networks in different areas of the brain (for examples, see Therrien et al⁴ and Slachevsky et al⁵). Importantly, this means the amount each contributes to a sustained change in movement control can be manipulated and reweighted by altering the task structure. This multiplicity of mechanisms for motor learning might be a powerful tool for physical therapists across areas of practice, but we must first understand the rules governing how each of them operates.

A key property of these motor learning mechanisms is that they are not necessarily yoked or dependent on one another. Although each mechanism requires that multiple movement trials or repetitions be completed, the neural process underlying the changes in movement during and after practice differ depending on the structure and the amount of task practice. More specifically, use-dependent learning and sensorimotor adaptation motor learning can operate simultaneously and one can be preferentially targeted without engaging the other.² Similar evidence exists for the independence of instructive^1,6,7 and reinforcement learning⁴ from sensorimotor adaptation and use-dependent learning. This is why we displayed the motor learning mechanisms to be parallel to one another in Figure 2 of the original manuscript. These distinct mechanisms operate in parallel (versus serially), which means that (1) each can be individually targeted and (2) if one is impaired (in the case of a neurologic injury or disease) the others are still available to enable motor learning. That said, further research is needed to better understand how these distinct, parallel learning mechanisms might interact during motor learning.^8,9

Macpherson et al also point out that our paper does not directly cover the application of this knowledge to neurologic patient populations. The choice not to include this topic in the paper was intentional for the following reasons. First, we wanted to keep the topic broadly applicable to multiple practice areas of physical therapy practice (eg, orthopedics, neurologic, pelvic health, etc). As such, a nuanced discussion about the application of these concepts to different neurologic populations was beyond the scope of the original article. Second, there is not yet enough evidence to indicate how disorders of the nervous system might impact the translation of these principles into practice.

Nevertheless, Macpherson et al highlight an important line of inquiry that deserves more attention—should a motor learning mechanism that relies on a part of the brain that is damaged be intentionally bypassed? Or, should that form of motor learning be incorporated to promote plasticity in that brain area? In the case of Parkinson disease (PD) that Macpherson et al discuss, reduced dopaminergic signaling likely results in less responsivity to rewards.¹⁰ Thus, patients with PD have an impaired or reduced capacity for reinforcement or reward-based motor learning, so targeting other motor learning mechanisms should be prioritized to promote changes in behavior. There is no direct evidence to support an intuition that the capacity to learn through reinforcement changes through the progression of PD. However, studies of sensorimotor adaptation in patients with cerebellar ataxia demonstrate that the reduced capacity for sensorimotor adaptation-based motor learning¹¹ is related to the severity of ataxia.¹² Importantly, people with cerebellar ataxia still exhibit the capacity to learn through other mechanisms (reinforcement^4,13 and use-dependent learning¹⁴). For other neurologic conditions, it might be more appropriate to target the damaged area of the central nervous system to promote plasticity (eg, spinal cord injury). However, much more research is needed to clarify how and when to optimally leverage these motor learning mechanisms in patients with different neurologic conditions.

Macpherson et al argue that we overgeneralized the results from experimental studies of tasks with limited ecological validity. We disagree. We were careful to emphasize that “further research is needed to determine how each of these distinct motor learning mechanisms can be effectively integrated into the design of functional mobility interventions.” Indeed, we agree that there is a dearth of well-designed motor learning studies that examine the kinds of complex tasks that patients face in their daily lives. The principal reason for this is that studies that can effectively tease apart the rules governing the different mechanisms of motor learning require over-specified task conditions that limit the options of the experimental subjects. Science must proceed carefully from such well-controlled studies to test whether the findings can be extended to different tasks and populations. Macpherson et al demonstrate this process nicely with their proposed translation of the experimental findings from van Nuland et al¹⁵ to guide the treatment of a patient with PD.

This section of the letter also warrants further clarification about motor learning tasks and the learning mechanisms they are designed to target. Embedded within their argument about the ecological validity of experimental motor learning tasks is a fundamental misunderstanding about these motor learning mechanisms and their distinct behavioral drivers. This is demonstrated in the following statement, “These studies rely on unusual perturbation-based tasks (ie, force fields, split-belt treadmill, prism goggles) and target novel skill learning…” Motor learning studies that use perturbation-based tasks are typically studies of sensorimotor adaptation-based learning (ie, recalibration of an existing motor program), not novel skill learning (ie, development of a wholly new motor program). During sensorimotor adaptation, an already learned existing motor program is modified in order to reduce a sensory prediction error. For example, when experimenters use force fields to study motor adaptation, the force field is introduced within the context of a well-practiced, familiar movement (eg, forward reaching). Novel skill learning, on the other hand, is studied by introducing someone to a motor task they have never been exposed to before.¹⁶ Often, these are contrived computer-based tasks, but in a recent publication by Bayani et al,¹⁷ an antiquated stone tool–making method called stone knapping was cleverly employed to examine the dynamics of novel motor skill learning during an anthropologically and ecologically valid task.

Macpherson et al contend that our “discussion of strategies to facilitate motor learning is incomplete and potentially misleading.” In our paper, we focused on summarizing 4 different motor learning mechanisms and the methods through which one can increase the relative contribution of each to an overall change in behavior. We omitted an in-depth discussion of general strategies to facilitate motor learning, because our paper focused on the mechanisms underlying motor learning and such strategies have previously been provided by others. We disagree that this was misleading, as we reference this work in the third paragraph of the paper. Furthermore, as we described in the “Future Research Directions” section, not much is known about how to optimize learning through each of these distinct motor learning mechanisms. Limited studies suggest that the same optimization principles might not apply to each process.¹⁸ For example, an external focus of attention might influence instructive motor learning, but it is unlikely to have an impact on sensorimotor adaptation.

We hope that this response serves to (1) clarify how these motor learning mechanisms operate to change motor behaviors; (2) highlight the strengths and limitations of the current motor learning research; (3) reinforce how these advancements in research might be used to inform clinical physical therapist practice; and (4) emphasize the opportunities for further contributions to this field that would be meaningful to the field of physical therapy. Again, we are grateful to Macpherson et al for their thoughtful critique and to the editors for providing the opportunity to broaden the conversation about this important topic.