In the last decades there has been an increasing number of studies investigating how the Central Nervous System (CNS) controls movements, by using robotic technology. Force field paradigms are able to guide internal model formation, i.e. sensorimotor adaptation mechanisms, as form of implicit learning [1]. In particular, multijoint reaching movements have been studied by applying robot-generated deviating forces [2]. Through haptic robots, indeed, the subjects are exposed to extrinsic dynamics (environment), whose effect is added to the effect of the intrinsic one [3–5]. This non-invasive tool allows to accurately observe and quantify the motor response of the CNS to an external dynamic interaction, investigating possible short-term adaptation strategies.

Patton and colleagues [6] exposed stroke patients to robot-generated force fields; they found out that significant improvements occurred only when the training forces magnified the original errors (error-enhancing therapy). A similar approach was employed by Masia and colleagues [7] on cerebral palsy children, with the final goal to identify a tool able to evaluate to which extent their motor learning capability is impaired.

The force field paradigm application on healthy subjects allowed to point out relevant features of motor learning; for instance, differences in motor mechanism tuning between children and adults emerged [8] and a kinematic variability was found as subjects move in different directions while experiencing the same structured force field [9, 10], due to the anisotropic features of the arm impedance.

In this framework, the response to altered dynamic environments and the evaluation of motor learning of children affected by primary dystonia have never been investigated. Dystonia is a syndrome characterized by excessive and sustained muscle contraction causing twisting and repetitive movements, abnormal postures, or both [11–13]. New insights suggest that dystonia could also have as origin a lack of reliable sensory feedbacks about motor actions [14], in particular an impairment of sensory integration of afferent inputs, such as proprioceptive signals [15, 16]. The ability of dystonic children to deal with the central planning issues associated with the control of arm motion is still an open question [17].

The purpose of the present study is to evaluate whether the application of force to limbs can improve motor learning in children affected by primary dystonia. Indeed, it is well-known that healthy adults can quickly adapt to novel dynamic environments, forming a mapping (an internal model) between limb state and muscle forces; it is less asserted a standard behavior concerning children.

Thus, explorative experimental sessions switching on/off an external force field, on healthy and dystonic children, performing motor tasks with different direction features, allow to outline the adaptation mechanisms and the error signals that drive these processes.

In terms of clinical fall-outs, this investigation explores the possibility of teaching to dystonic patients desired movements using after-effects from adaptation to robot-applied force fields, in particular exploiting the error-amplifying forces, whose potential is justified by the observation that movement error is likely to be a driving signal for adaptation and learning [18, 19]. All these insights would suggest whether and how adaptive training could provide an effective supplement to conventional interventions, which can be of limited efficacy, for dystonia. Indeed, the current treatment options include pharmacological and surgical strategies; in children with primary dystonia, which is often generalized, the results may be not satisfactory, not steady in long-term and not effective for all tasks, hence specific rehabilitative trainings are needed for better improvements. Which neurorehabilitative exercises are the most effective for dystonic people is still matter of discussion [20].

This paper explores the features of motor adaptation in children with primary dystonia during the execution of multijoint movements that are disturbed by a force field, comparing their behavior with an aged-matched control group. Both groups completed the required tasks. However, there are clear evidences of different performances between impaired and unimpaired children.

Our findings from healthy subjects show that the disturbing force affects significantly the movement outcomes. Moreover, while the dynamic alteration is acting, a complete compensation is not achieved, i.e. the trajectory does not return to be as straight as in the reference condition. Fifteen repetitions, not consecutive but inserted into a direction-change sequence for the reaching task, do not allow to complete the updating of the internal model of body/environment which drives the feedforward control mechanism. Only in upward reaching a tendency to carry out compensatory strategies, canceling the constant perturbation effect on the trajectory, emerged.

The absence of systematic adaptation during the divergent force field in healthy children could be better investigated in order to understand how much it is ascribable to the defined timing of the protocol phases or to the young age, since it is known that, to optimize motor learning, children may require longer periods of practice than adults [8, 23].

Alternatively, it is possible that the motor behavior is not recovered as a matching with initial reference performance (A condition). According to the theory proposed by Izawa and colleagues [24], the motor control in a novel environment is not a process of perturbation cancellation, rather, the process resembles reoptimization. Thus, it would support the hypothesis that the control of action proceeds via two related pathways: on the one hand, adaptation produces a more accurate estimate of the sensory consequences of the motor commands (i.e. learning an accurate forward model), and on the other hand, the brain searches for a better movement plan which minimizes an implicit motor cost and maximizes rewards (i.e. finding an optimum controller).

Anyway, the lack of adaptation in B condition is consistent with the lack of after-effects in C condition, when the disturbance is re-deleted; indeed, the motor control when the dynamic alteration is not acting anymore is perfectly comparable with the reference one. Also Burdet and colleagues [25] found for healthy subjects performing reaching that, when the force field was removed, the trajectories were even straighter than in the reference condition: after-effects were absent following adaptation to a destabilizing force field that amplified trajectory errors. It was associated to the CNS skill of tuning impedance, in order to achieve stability.

Thus, for healthy group, the disturbance uncalibrates the sensorimotor system. It is known that internal models, i.e. body/environment representations, are efficient for motor control only if they produce unbiased predictions of body states; it requires that the level of noise in the system is sufficiently low and mainly that the sensorimotor system is well calibrated [26, 27].

In the light of what we observed about the healthy controls, the dystonic subjects show a less efficient internal model, regardless the external disturbance. In fact, in the reference condition without any disturbing factors, they carry out a less precise, more variable and less reliable path control than the healthy ones. Since their sensorimotor system is already not-well calibrated, the disturbance does not systematically affect their motor control. It has been shown that dystonia is characterized by abnormal sensory integration, i.e. by an incomplete processing of the incoming signals, resulting in distorted information and thus in abnormal motor outputs [15].

However, it is particular intriguing how the altered dynamic exposure (B condition) would induce a subsequent improvement (C condition), in terms of optimal path control, in dystonic population, despite a lack of clear adaptation during the exposure to the constant disturbance.

Rossetti and colleagues [27] speculated about the mechanisms that might allow a stroke subject to decrease directional errors with force field paradigm based on distorting interventions, that cannot decrease by simple practice alone: sensory feedback systems may need to detect a stimulus with a magnitude that is large enough to trigger the recovery process [28–30]. Many of the substrates related to robot-mediated motor adaptation overlap with brain areas related to motor recovery after a CNS injury such as stroke [31, 32]. Masia and colleagues [7], interpreting such mechanisms in children with cerebral palsy, proposed another explanation: the nervous system is trying to use motor pathways that are no longer intact, and the learning is a way to trick the nervous system into a new and non-intuitive pathway that it would otherwise not ever consider; however, such ideas would need movement-correlated brain imaging studies to be validated [33, 34]. Speculating on our findings in the same framework, we transfer these considerations on dystonia; the external sensory element could induce a “motor control improvement” by enhancing the sensorimotor system calibration. In other words, a very short-term environment alteration is probably not able to establish new sensory engrams for the new dynamic environment (i.e. specific learned and memorized motor patterns stored in both sensory and motor portions of the brain), but it is able to refine the sub-optimal standard sensorimotor patterns for a specific task. The suggestion is that the disturbance could not induce single trials evolution, but a global effect triggered by the just-experienced dynamic alteration. The force-field paradigm induces adaptation in a relatively short timescale, on the order of tens of movements, which makes it possible to perform motor-learning experiments quickly [35].

The uncalibrated dystonic sensorimotor system seems to have improvement margins: the force alterations would induce a more effective recruitment of cortico-cortical connections linking the ipsilateral motor and somatosensory cortical areas. The short-time error-enhancing therapy in dystonia could represent a training to refine the existing but strongly imprecise motor scheme.

As for healthy, the behaviour in C condition for dystonic children does not show after-effects. Differently from the healthy CNS which can control impedance to enhance the robusteness to external perturbations or to biased sensorimotor transformations [36], dystonic subjects could not be able to control the full impedance, which would mean to minimize the metabolic cost [25]; they likely act only on stiffness control, which is realized by prolonging the movement duration and/or co-contracting muscular activity, thus causing bradykinesia and increased motor output variability [12, 37]. Trial-to-trial variability indeed arises from neural sources [38, 39] and is larger during childhood [40] or due to a brain damage.

These underlying mechanisms, both in healthy and in dystonic subjects, should be better investigated also in relation to the directionality of motor learning, about which here some cues emerged from reaching tasks. For more robust conclusions, different force field paradigms (e.g. different directions and intensities), even subject-specific, should be tested and wider populations with systematic features, possibly untreated patients, should be recruited.

The main limitation of this work is the quite small number of trials; this choice in the protocol definition arose by the need to not introduce fatigue and demotivation which could confuse the comparisons between the consecutive conditions. Moreover, most of the dystonic children were under treatment (medical and/or deep brain stimulation) and they displayed a great variability in the severity of dystonia, even if all affected homogeneously by primary dystonia. Another major issue to be addressed in details in the next related studies is the persistence of such potential beneficial effects after a sequence of training sessions exposing the patient to dynamic alterations and waiting for a washout period.

The basic scientific knowledge gained with the robotic force-field paradigm most likely will lead to practical enhancements in motor learning issues. A thorough understanding of the error signals that drive adaptation may allow them to be amplified or filtered, thus accelerating learning.

In conclusion, this work highlights encouraging evidence that haptic training could provide an effective supplement to conventional therapy in dystonia. Thus, the neural processes associated with implicit motor adaptation may reshape sensorimotor mappings altered by dystonia that cannot be tuned simply by practicing movement.