Adaptation and generalization in acceleration-dependent force fields

Dissociating the Influence of Limb Posture and Visual Feedback Shifts on the Adaptation to Novel Movement Dynamics

Accurate movements of the upper limb require the integration of various forms of sensory feedback (e.g., visual and postural information). The influence of these different sensory modalities on reaching movements has been largely studied by assessing endpoint errors after selectively perturbing sensory estimates of hand location. These studies have demonstrated that both vision and proprioception make key contributions in determining the reach endpoint. However, their influence on motor output throughout movement remains unclear. Here we used separate perturbations of posture and visual information to dissociate their effects on reaching dynamics and temporal force profiles during point-to-point reaching movements. We tested human subjects (N = 32) and found that vision and posture modulate select aspects of reaching dynamics. Specifically, altering arm posture influences the relationship between temporal force patterns and the motion-state variables of hand position and acceleration, whereas dissociating visual feedback influences the relationship between force patterns and the motion-state variables of velocity and acceleration. Next, we examined the extent these baseline motion-state relationships influence motor adaptation based on perturbations of movement dynamics. We trained subjects using a velocity-dependent force-field to probe the extent arm posture-dependent influences persisted after exposure to a motion-state dependent perturbation. Changes in the temporal force profiles due to variations in arm posture were not reduced by adaptation to novel movement dynamics, but persisted throughout learning. These results suggest that vision and posture differentially influence the internal estimation of limb state throughout movement and play distinct roles in forming the response to external perturbations during movement.

Inferring control objectives in a virtual balancing task in humans and monkeys

Natural behaviors have redundancy, which implies that humans and animals can achieve their goals with different strategies. Given only observations of behavior, is it possible to infer the control objective that the subject is employing? This challenge is particularly acute in animal behavior because we cannot ask or instruct the subject to use a particular strategy. This study presents a three-pronged approach to infer an animal's control objective from behavior. First, both humans and monkeys performed a virtual balancing task for which different control strategies could be utilized. Under matched experimental conditions, corresponding behaviors were observed in humans and monkeys. Second, a generative model was developed that represented two main control objectives to achieve the task goal. Model simulations were used to identify aspects of behavior that could distinguish which control objective was being used. Third, these behavioral signatures allowed us to infer the control objective used by human subjects who had been instructed to use one control objective or the other. Based on this validation, we could then infer objectives from animal subjects. Being able to positively identify a subject's control objective from observed behavior can provide a powerful tool to neurophysiologists as they seek the neural mechanisms of sensorimotor coordination.

Dorsomedial frontal cortex damage impairs error-based, but not reinforcement-based motor learning in humans

We adapt our movements to new and changing environments through multiple processes. Sensory error-based learning counteracts environmental perturbations that affect the sensory consequences of movements. Sensory errors also cause the upregulation of reflexes and muscle co-contraction. Reinforcement-based learning enhances the selection of movements that produce rewarding outcomes. Although some findings have identified dissociable neural substrates of sensory error- and reinforcement-based learning, correlative methods have implicated dorsomedial frontal cortex in both. Here, we tested the causal contributions of dorsomedial frontal to adaptive motor control, studying people with chronic damage to this region. Seven human participants with focal brain lesions affecting the dorsomedial frontal and 20 controls performed a battery of arm movement tasks. Three experiments tested: (i) the upregulation of visuomotor reflexes and muscle co-contraction in response to unpredictable mechanical perturbations, (ii) sensory error-based learning in which participants learned to compensate predictively for mechanical force-field perturbations, and (iii) reinforcement-based motor learning based on binary feedback in the absence of sensory error feedback. Participants with dorsomedial frontal damage were impaired in the early stages of force field adaptation, but performed similarly to controls in all other measures. These results provide evidence for a specific and selective causal role for the dorsomedial frontal in sensory error-based learning.

Inferring control objectives in a virtual balancing task in humans and monkeys

Natural behaviors have redundancy, which implies that humans and animals can achieve their goals with different control objectives. Given only observations of behavior, is it possible to infer the control strategy that the subject is employing? This challenge is particularly acute in animal behavior because we cannot ask or instruct the subject to use a particular control strategy. This study presents a threepronged approach to infer an animal's control strategy from behavior. First, both humans and monkeys performed a virtual balancing task for which different control objectives could be utilized. Under matched experimental conditions, corresponding behaviors were observed in humans and monkeys. Second, a generative model was developed that represented two main control strategies to achieve the task goal. Model simulations were used to identify aspects of behavior that could distinguish which control objective was being used. Third, these behavioral signatures allowed us to infer the control objective used by human subjects who had been instructed to use one control objective or the other. Based on this validation, we could then infer strategies from animal subjects. Being able to positively identify a subject's control objective from behavior can provide a powerful tool to neurophysiologists as they seek the neural mechanisms of sensorimotor coordination.

Inconsistent attentional contexts impair re-learning following gradual visuomotor adaptation

One of the brain's primary functions is to promote actions in dynamic, distracting environments. Because distractions divert attention away from our primary goals, we learn to maintain accurate actions under sensory and cognitive distractions. Visuomotor adaptation refers to learning processes that restore performance when sensorimotor capacities or environmental conditions are abruptly or gradually altered. Prior work showed that learning to counteract an abrupt perturbation while performing either a single or dual task, referred to as the attentional context, led to better and faster re-learning when the same attentional context was reinstated at recall. This suggested that the attentional context was associated with visuomotor adaptation and used as a contextual cue during recall. The current study investigated whether attentional context was associated with visuomotor adaptation to a gradual perturbation, which limits awareness of errors. During adaptation, participants reached to targets while learning to counteract a visuomotor rotation that increased from 0 to 45 deg by 0.3 deg each trial, with or without performing a secondary task. Re-learning was impaired when the attentional context changed between adaptation and recall (Experiment 1), even compared to when the secondary task was only performed during the early or late half of adaptation (Experiment 2). Changing the secondary task between adaptation and recall did not impair re-learning, indicating that the effect was attentional-context-dependent, rather than task-specific (Experiment 3). These findings further highlight the importance of cognitive factors, such as attention, in visuomotor adaptation and have implications for learning and rehabilitation paradigms.

The speed of adaptation is dependent on the load type during target reaching by intact human subjects

When lifting or moving a novel object, humans are routinely able to quickly characterize the nature of the unknown load and swiftly achieve the desired movement trajectory. It appears that both tactile and proprioceptive feedback systems help humans develop an accurate prediction of load properties and determine how associated limb segments behave during voluntary movements. While various types of limb movement information, such as position, velocity, acceleration, and manipulating forces, can be detected using human tactile and proprioceptive systems, we know little about how the central nervous system decodes these various types of movement data, and in which order or priority they are used when developing predictions of joint motion during novel object manipulation. In this study, we tested whether the ability to predict motion is different between position- (elastic), velocity- (viscous), and acceleration-dependent (inertial) loads imposed using a multiaxial haptic robot. Using this protocol, we can learn if the prediction of the motion model is optimized for one or more of these types of mechanical load. We examined ten neurologically intact subjects. Our key findings indicated that inertial and viscous loads showed the fastest adaptation speed, whereas elastic loads showed the slowest adaptation speed. Different speeds of adaptation were observed across different magnitudes of the load, suggesting that human capabilities for predicting joint motion and manipulating loads may vary systematically with different load types and load magnitudes. Our results imply that human capabilities for load manipulation seems to be most sensitive to and potentially optimized for inertial loads.

On the encoding capacity of human motor adaptation

Primitive-based models of motor learning suggest that adaptation occurs by tuning the responses of motor primitives. Based on this idea, we consider motor learning as an information encoding procedure, that is, a procedure of encoding a motor skill into primitives. The capacity of encoding is determined by the number of recruited primitives, which depends on how many primitives are "visited" by the movement, and this leads to a rather counterintuitive prediction that faster movement, where a larger number of motor primitives are involved, allows learning more complicated motor skills. Here, we provide a set of experimental results that support this hypothesis. First, we show that learning occurs only with movement, that is, only with nonzero encoding capacity. When participants were asked to counteract a rotating force applied to a robotic handle, they were unable to do so when maintaining a static posture but were able to adapt when making small circular movements. Our second experiment further investigated how adaptation is affected by movement speed. When adapting to a simple (low-information-content) force field, fast (high-capacity) movement did not have an advantage over slow (low-capacity) movement. However, for a complex (high-information-content) force field, the fast movement showed a significant advantage over slow movement. Our final experiment confirmed that the observed benefit of high-speed movement is only weakly affected by mechanical factors. Taken together, our results suggest that the encoding capacity is a genuine limiting factor of human motor adaptation.NEW & NOTEWORTHY We propose a novel concept called "encoding capacity" of motor adaptation, which describes an inherent limiting-factor of our brain's ability to learn new motor skills, just like any other storage system. By reinterpreting the existing primitive-based models of motor learning, we hypothesize that the encoding capacity is determined by the size of the movement, and present a set of experimental evidence suggesting that such limiting effect of encoding capacity does exist in human motor adaptation.

Force field generalization and the internal representation of motor learning

When learning a new motor behavior, e.g. reaching in a force field, the nervous system builds an internal representation. Examining how subsequent reaches in unpracticed directions generalize reveals this representation. Although often studied, it is not known how this representation changes across training directions, or how changes in reach direction and the corresponding changes in limb impedance, influence these measurements. We ran a force field adaptation experiment using eight groups of subjects each trained on one of eight standard directions and then tested for generalization in the remaining seven directions. Generalization in all directions was local and asymmetric, providing limited and unequal transfer to the left and right side of the trained target. These asymmetries were not consistent in either magnitude or direction, even after correcting for changes in limb impedance. Relying on a standard model for generalization the inferred representations inconsistently shifted to one side or the other of their respective training direction. A second model that accounted for limb impedance and variations in baseline trajectories explained more data and the inferred representations were centered on their respective training directions. Our results highlight the influence of limb mechanics and impedance on psychophysical measurements and their interpretations for motor learning.

A computational scheme for internal models not requiring precise system parameters

Utilization by humans of a precise and adaptable internal model of the dynamics of the body in generating movements is a well-supported concept. The prevailing opinion is that such an internal model ceaselessly develops through long-term repetition and accumulation in the central nervous system (CNS). However, a long-term learning process would not be absolutely necessary for the formation of internal models. It is possible to estimate the dynamics of the system by using a motor command and its resulting output, instead of constructing a model of the dynamics with precise parameters. In this study, a computational model is proposed that uses a motor command and its corresponding output to estimate the dynamics of the system and it is examined whether the proposed model is capable of describing a series of empirical movements. The proposed model was found to be capable of describing humans' fast movements which require compensation for system dynamics as well as sensory delays. In addition, the proposed model shows equifinality under inertial perturbations as seen in several experimental studies. This satisfactory reproducibility of the proposed computation raises the possibility that humans make a movement by estimating the system dynamics with a copy of motor command and sensory output on a momentary basis, without the need to identify precise system parameters.

Delineating cerebellar mechanisms in DYT11 myoclonus-dystonia

BACKGROUND

Recent research has highlighted the role of the cerebellum in the pathophysiology of myoclonus-dystonia syndrome as a result of mutations in the ɛ-sarcoglycan gene (DYT11). Specifically, a cerebellar-dependent saccadic adaptation task is dramatically impaired in this patient group.

OBJECTIVES

The objective of this study was to investigate whether saccadic deficits coexist with impairments of limb adaptation to provide a potential mechanism linking cerebellar dysfunction to the movement disorder within symptomatic body regions.

METHODS

Limb adaptation to visuomotor (visual feedback rotated by 30°) and forcefield (force applied by robot to deviate arm) perturbations were examined in 5 patients with DYT11 and 10 aged-matched controls.

RESULTS

Patients with DYT11 successfully adapted to both types of perturbation. Modelled and averaged summary metrics that captured adaptation behaviors were equivalent to the control group across conditions.

CONCLUSIONS

DYT11 is not characterized by a uniform deficit in adaptation. The previously observed large deficit in saccadic adaption is not reflected in an equivalent deficit in limb adaptation in symptomatic body regions. We suggest potential mechanisms at the root of this discordance and identify key research questions that need future study. © 2018 International Parkinson and Movement Disorder Society.

Direct-effects and after-effects of dynamic adaptation on intralimb and interlimb transfer

After-effects following sensorimotor adaptation are generally considered as evidence for the formation of an internal model, although evidence lacks on whether the absence of after-effects necessarily indicates that the adaptation did not result in the formation of an internal model. Here, we examined direct- and after-effects of dynamic adaptation with one arm at one workspace on subsequent performance with the other arm, as well as the same arm at another workspace. During training, subjects performed reaching movements under a novel dynamic condition with the right arm; during testing, they performed reaching movements with the left or right arm at a new workspace, under either the same dynamic condition (direct-effects) or a normal condition (after-effects). Results showed significant transfer within the same arm in terms of both direct- and after-effects, but significant transfer across the arms only in terms of direct-effects. These findings suggest that the formation of an internal model does not always result in after-effects. They also support the idea that the neural representation developed after sensorimotor adaptation comprise some aspects that are effector independent and other aspects that are effector dependent; and that direct- and after-effects following sensorimotor adaptation mainly reflect the effector-independent and the effector-dependent aspects, respectively.

Learning to Predict and Control the Physics of Our Movements

When we hold an object in our hand, the mass of the object alters the physics of our arm, changing the relationship between motor commands that our brain sends to our arm muscles and the resulting motion of our hand. If the object is unfamiliar to us, our first movement will exhibit an error, producing a trajectory that is different from the one we had intended. This experience of error initiates learning in our brain, making it so that on the very next attempt our motor commands partially compensate for the unfamiliar physics, resulting in smaller errors. With further practice, the compensation becomes more complete, and our brain forms a model that predicts the physics of the object. This model is a motor memory that frees us from having to relearn the physics the next time that we encounter the object. The mechanism by which the brain transforms sensory prediction errors into corrective motor commands is the basis for how we learn the physics of objects with which we interact. The cerebellum and the motor cortex appear to be critical for our ability to learn physics, allowing us to use tools that extend our capabilities, making us masters of our environment.

The temporal stability of visuomotor adaptation generalization

Movement adaptation in response to systematic motor perturbations exhibits distinct spatial and temporal properties. These characteristics are typically studied in isolation, leaving the interaction largely unknown. Here we examined how the temporal decay of visuomotor adaptation influences the spatial generalization of the motor recalibration. First, we quantified the extent to which adaptation decayed over time. Subjects reached to a peripheral target, and a rotation was applied to the visual feedback of the unseen motion. The retention of this adaptation over different delays (0-120 s) 1) decreased by 29.0 ± 6.8% at the longest delay and 2) was represented by a simple exponential, with a time constant of 22.5 ± 5.6 s. On the basis of this relationship we simulated how the spatial generalization of adaptation would change with delay. To test this directly, we trained additional subjects with the same perturbation and assessed transfer to 19 different locations (spaced 15° apart, symmetric around the trained location) and examined three delays (~4, 12, and 25 s). Consistent with the simulation, we found that generalization around the trained direction (±15°) significantly decreased with delay and distance, while locations >60° displayed near-constant spatiotemporal transfer. Intermediate distances (30° and 45°) showed a difference in transfer across space, but this amount was approximately constant across time. Interestingly, the decay at the trained direction was faster than that based purely on time, suggesting that the spatial transfer of adaptation is modified by concurrent passive (time dependent) and active (movement dependent) processes.NEW & NOTEWORTHY Short-term motor adaptation exhibits distinct spatial and temporal characteristics. Here we investigated the interaction of these features, utilizing a simple motor adaptation paradigm (recalibration of reaching arm movements in response to rotated visual feedback). We examined the changes in the spatial generalization of motor adaptation for different temporal manipulations and report that the spatiotemporal generalization of motor adaptation is generally local and is influenced by both passive (time dependent) and active (movement dependent) learning processes.

Representing delayed force feedback as a combination of current and delayed states

To adapt to deterministic force perturbations that depend on the current state of the hand, internal representations are formed to capture the relationships between forces experienced and motion. However, information from multiple modalities travels at different rates, resulting in intermodal delays that require compensation for these internal representations to develop. To understand how these delays are represented by the brain, we presented participants with delayed velocity-dependent force fields, i.e., forces that depend on hand velocity either 70 or 100 ms beforehand. We probed the internal representation of these delayed forces by examining the forces the participants applied to cope with the perturbations. The findings showed that for both delayed forces, the best model of internal representation consisted of a delayed velocity and current position and velocity. We show that participants relied initially on the current state, but with adaptation, the contribution of the delayed representation to adaptation increased. After adaptation, when the participants were asked to make movements with a higher velocity for which they had not previously experienced with the delayed force field, they applied forces that were consistent with current position and velocity as well as delayed velocity representations. This suggests that the sensorimotor system represents delayed force feedback using current and delayed state information and that it uses this representation when generalizing to faster movements.NEW & NOTEWORTHY The brain compensates for forces in the body and the environment to control movements, but it is unclear how it does so given the inherent delays in information transmission and processing. We examined how participants cope with delayed forces that depend on their arm velocity 70 or 100 ms beforehand. After adaptation, participants applied opposing forces that revealed a partially correct representation of the perturbation using the current and the delayed information.

Temporal specificity of the initial adaptive response in motor adaptation

Repeated exposure to a novel physical environment eventually leads to a mature adaptive response whereby feedforward changes in motor output mirror both the amplitude and temporal structure of the environmental perturbations. However, adaptive responses at the earliest stages of learning have been found to be not only smaller, but systematically less specific in their temporal structure compared to later stages of learning. This observation has spawned a lively debate as to whether the temporal structure of the initial adaptive response is, in fact, stereotyped and non-specific. To settle this debate, we directly measured the adaptive responses to velocity-dependent and position-dependent force-field perturbations (vFFs and pFFs) at the earliest possible stage of motor learning in humans–after just a single-movement exposure. In line with previous work, we found these earliest stage adaptive responses to be more similar than the perturbations that induced them. However, the single-trial adaptive responses for vFF and pFF perturbations were clearly distinct, and the disparity between them reflected the difference between the temporal structure of the perturbations that drove them. Critically, we observed these differences between single-trial adaptive responses when vFF and pFF perturbations were randomly intermingled from one trial to the next within the same block, indicating perturbation response specificity at the single trial level. These findings demonstrate that the initial adaptive responses to physical perturbations are not stereotyped. Instead, the neural plasticity in sensorimotor areas is sensitive to the temporal structure of a movement perturbation even at the earliest stage in learning. This insight has direct implications for the development of computational models of early-stage motor adaptation and the evolution of this adaptive response with continued training.

With repeated exposure to a perturbation, the sensorimotor system learns to develop an adaptive response that is highly specific to both the amplitude and temporal structure of that perturbation in order to effectively counteract it. It is widely known that the amplitude of the adaptive response starts small and gradually grows to the right size with repeated exposure. However, it is also the case that the temporal structure of the adaptive response starts somewhat generically and gradually grows into the right shape with repeated exposure. A key question is whether the adaptive response to a perturbation begins with a stereotyped temporal structure that only becomes specified with further practice, or if it begins with a degree of specificity for the experienced perturbation that need only to be refined by practice. Here, by precisely measuring the temporal pattern of motor output in the single-trial adaptive response to two different perturbations, we show that the initial adaptive response is indeed specific to the temporal characteristics of the perturbation, even when the disturbance randomly changed from one trial to the next. These results demonstrate that the sensorimotor system is sensitive to the temporal features of a disturbance, even when experienced just once.

Transfer of dynamic motor skills acquired during isometric training to free motion

Recent studies have explored the prospects of learning to move without moving, by displaying virtual arm movement related to exerted force. However, it has yet to be tested whether learning the dynamics of moving can transfer to the corresponding movement. Here we present a series of experiments that investigate this isometric training paradigm. Subjects were asked to hold a handle and generate forces as their arms were constrained to a static position. A precise simulation of reaching was used to make a graphic rendering of an arm moving realistically in response to the measured interaction forces and simulated environmental forces. Such graphic rendering was displayed on a horizontal display that blocked their view to their actual (statically constrained) arm and encouraged them to believe they were moving. We studied adaptation of horizontal, planar, goal-directed arm movements in a velocity-dependent force field. Our results show that individuals can learn to compensate for such a force field in a virtual environment and transfer their new skills to the actual free motion condition, with performance comparable to practice while moving. Such nonmoving techniques should impact various training conditions when moving may not be possible.NEW & NOTEWORTHY This study provided early evidence supporting that training movement skills without moving is possible. In contrast to previous studies, our study involves 1) exploiting cross-modal sensory interactions between vision and proprioception in a motionless setting to teach motor skills that could be transferable to a corresponding physical task, and 2) evaluates the movement skill of controlling muscle-generated forces to execute arm movements in the presence of external forces that were only virtually present during training.

Effect of Position- and Velocity-Dependent Forces on Reaching Movements at Different Speeds

The speed of voluntary movements is determined by the conflicting needs of maximizing accuracy and minimizing mechanical effort. Dynamic perturbations, e.g., force fields, may be used to manipulate movements in order to investigate these mechanisms. Here, we focus on how the presence of position- and velocity-dependent force fields affects the relation between speed and accuracy during hand reaching movements. Participants were instructed to perform reaching movements under visual control in two directions, corresponding to either low or high arm inertia. The subjects were required to maintain four different movement durations (very slow, slow, fast, very fast). The experimental protocol included three phases: (i) familiarization—the robot generated no force; (ii) force field—the robot generated a force; and (iii) after-effect—again, no force. Participants were randomly assigned to four groups, depending on the type of force that was applied during the “force field” phase. The robot was programmed to generate position-dependent forces—with positive (K+) or negative stiffness (K−)—or velocity-dependent forces, with either positive (B+) or negative viscosity (B−). We focused on path curvature, smoothness, and endpoint error; in the latter we distinguished between bias and variability components. Movements in the high-inertia direction are smoother and less curved; smoothness also increases with movement speed. Endpoint bias and variability are greater in, respectively, the high and low inertia directions. A robust dependence on movement speed was only observed in the longitudinal components of both bias and variability. The strongest and more consistent effects of perturbation were observed with negative viscosity (B−), which resulted in increased variability during force field adaptation and in a reduction of the endpoint bias, which was retained in the subsequent after-effect phase. These findings confirm that training with negative viscosity produces lasting effects in movement accuracy at all speeds.

Human muscle spindle sensitivity reflects the balance of activity between antagonistic muscles

Muscle spindles are commonly considered as stretch receptors encoding movement, but the functional consequence of their efferent control has remained unclear. The "α-γ coactivation" hypothesis states that activity in a muscle is positively related to the output of its spindle afferents. However, in addition to the above, possible reciprocal inhibition of spindle controllers entails a negative relationship between contractile activity in one muscle and spindle afferent output from its antagonist. By recording spindle afferent responses from alert humans using microneurography, I show that spindle output does reflect antagonistic muscle balance. Specifically, regardless of identical kinematic profiles across active finger movements, stretch of the loaded antagonist muscle (i.e., extensor) was accompanied by increased afferent firing rates from this muscle compared with the baseline case of no constant external load. In contrast, spindle firing rates from the stretching antagonist were lowest when the agonist muscle powering movement (i.e., flexor) acted against an additional resistive load. Stepwise regressions confirmed that instantaneous velocity, extensor, and flexor muscle activity had a significant effect on spindle afferent responses, with flexor activity having a negative effect. Therefore, the results indicate that, as consequence of their efferent control, spindle sensitivity (gain) to muscle stretch reflects the balance of activity between antagonistic muscles rather than only the activity of the spindle-bearing muscle.

Normal motor adaptation in cervical dystonia: a fundamental cerebellar computation is intact

The potential role of the cerebellum in the pathophysiology of dystonia has become a focus of recent research. However, direct evidence for a cerebellar contribution in humans with dystonia is difficult to obtain. We examined motor adaptation, a test of cerebellar function, in 20 subjects with primary cervical dystonia and an equal number of aged matched controls. Adaptation to both visuomotor (distorting visual feedback by 30°) and forcefield (applying a velocity-dependent force) conditions were tested. Our hypothesis was that cerebellar abnormalities observed in dystonia research would translate into deficits of cerebellar adaptation. We also examined the relationship between adaptation and dystonic head tremor as many primary tremor models implicate the cerebellothalamocortical network which is specifically tested by this motor paradigm. Rates of adaptation (learning) in cervical dystonia were identical to healthy controls in both visuomotor and forcefield tasks. Furthermore, the ability to adapt was not clearly related to clinical features of dystonic head tremor. We have shown that a key motor control function of the cerebellum is intact in the most common form of primary dystonia. These results have important implications for current anatomical models of the pathophysiology of dystonia. It is important to attempt to progress from general statements that implicate the cerebellum to a more specific evidence-based model. The role of the cerebellum in this enigmatic disease perhaps remains to be proven.

Motor adaptation and generalization of reaching movements using motor primitives based on spatial coordinates

The brain processes sensory and motor information in a wide range of coordinate systems, ranging from retinal coordinates in vision to body-centered coordinates in areas that control musculature. Here we focus on the coordinate system used in the motor cortex to guide actions and examine physiological and psychophysical evidence for an allocentric reference frame based on spatial coordinates. When the equations of motion governing reaching dynamics are expressed as spatial vectors, each term is a vector cross product between a limb-segment position and a velocity or acceleration. We extend this computational framework to motor adaptation, in which the cross-product terms form adaptive bases for canceling imposed perturbations. Coefficients of the velocity- and acceleration-dependent cross products are assumed to undergo plastic changes to compensate the force-field or visuomotor perturbations. Consistent with experimental findings, each of the cross products had a distinct reference frame, which predicted how an acquired remapping generalized to untrained location in the workspace. In response to force field or visual rotation, mainly the coefficients of the velocity- or acceleration-dependent cross products adapted, leading to transfer in an intrinsic or extrinsic reference frame, respectively. The model further predicted that remapping of visuomotor rotation should under- or overgeneralize in a distal or proximal workspace. The cross-product bases can explain the distinct patterns of generalization in visuomotor and force-field adaptation in a unified way, showing that kinematic and dynamic motor adaptation need not arise through separate neural substrates.

Two different motor learning mechanisms contribute to learning reaching movements in a rotated visual environment

Practice of movement in virtual-reality and other artificially altered environments has been proposed as a method for rehabilitation following neurological injury and for training new skills in healthy humans.  For such training to be useful, there must be transfer of learning from the artificial environment to the performance of desired skills in the natural environment.  Therefore an important assumption of such methods is that practice in the altered environment engages the same learning and plasticity mechanisms that are required for skill performance in the natural environment.  We test the hypothesis that transfer of learning may fail because the learning and plasticity mechanism that adapts to the altered environment is different from the learning mechanism required for improvement of motor skill.  In this paper, we propose that a model that separates skill learning and environmental adaptation is necessary to explain the learning and aftereffects that are observed in virtual reality experiments.  In particular, we studied the condition where practice in the altered environment should lead to correct skill performance in the original environment. Our 2-mechanism model predicts that aftereffects will still be observed when returning to the original environment, indicating a lack of skill transfer from the artificial environment to the original environment. To illustrate the model prediction, we tested 10 healthy participants on the interaction between a simple overlearned motor skill (straight hand movements to targets in different directions) and an artificially altered visuomotor environment (rotation of visual feedback of the results of movement).  As predicted by the models, participants show adaptation to the altered environment and after-effects on return to the baseline environment even when practice in the altered environment should have led to correct skill performance.  The presence of aftereffect under all conditions that involved changes in environment demonstrates separation of environmental adaptation and skill learning. Our results support the existence of two distinct learning modules with different adaptation properties.  Therefore we suggest that adaptation to an altered environment may not be useful for training new skills.

Vestibular benefits to task savings in motor adaptation

In everyday life, we seamlessly adapt our movements and consolidate them to multiple behavioral contexts. This natural flexibility seems to be contingent on the presence of movement-related sensorimotor cues and cannot be reproduced when static visual or haptic cues are given to signify different behavioral contexts. So far, only sensorimotor cues that dissociate the sensorimotor plans prior to force field exposure have been successful in learning two opposing perturbations. Here we show that vestibular cues, which are only available during the perturbation, improve the formation and recall of multiple control strategies. We exposed subjects to inertial forces by accelerating them laterally on a vestibular platform. The coupling between reaching movement (forward-backward) and acceleration direction (leftward-rightward) switched every 160 trials, resulting in two opposite force environments. When exposed for a second time to the same environment, with the opposite environment in between, subjects showed retention resulting in an ∼3 times faster adaptation rate compared with the first exposure. Our results suggest that vestibular cues provide contextual information throughout the reach, which is used to facilitate independent learning and recall of multiple motor memories. Vestibular cues provide feedback about the underlying cause of reach errors, thereby disambiguating the various task environments and reducing interference of motor memories.

State dependence of adaptation of force output following movement observation

Humans readily learn to move through direct physical practice and by watching the movements of others. Some researchers have proposed that action observation can inform subsequent changes in control through the acquisition of a neural representation of the novel dynamics, but to date learning following observation has been described by kinematic metrics. Here we designed an experiment to consider the specificity of adaptation to novel dynamic perturbations at the level of force generation. We measured changes in temporal patterns of force output following either the performance or observation of movements perturbed by either position- or velocity-dependent dynamic environments to 1) establish whether previously described observational motor learning effects were attributable to changes in predictive limb control and 2) determine whether such adaptation reflected a learned dependence on limb states appropriate to the haptic environment. We found that subjects who observed perturbed movements produced significant compensatory changes in their lateral force output, despite never directly experiencing force perturbations firsthand while performing the motor task. The time series of observers' adapted force outputs suggested that the state dependence of observed dynamics shapes adaptation. We conclude that the brain can transform observation of kinematics into state-dependent adaptation of reach dynamics.

Brevity of haptic force perturbations induces heightened adaptive sensitivity

We have exposed human participants to both full-movement and pulsatile viscous force perturbations to study the effect of force duration on the incremental transformation of sensation into adaptation. Traditional views of movement biomechanics could suggest that pulsatile forces would largely be attenuated as stiffness and viscosity act as a natural low-pass filter. Sensory transduction, however, tends to react to changes in stimuli and therefore could underlie heightened sensitivity to briefer, pulsatile forces. Here, participants adapted within perturbation duration conditions in a manner proportionate to sensed force and positional errors. Across perturbation conditions, we found participants had greater adaptive sensitivity when experiencing pulsatile forces rather than full-movement forces. In a follow-up experiment, we employed error-clamped, force channel trials to determine changes in predictive force generation. We found that while participants learned to closely compensate for the amplitude and breadth of full-movement forces, they exhibited a persistent mismatch in amplitude and breadth between adapted motor output and experienced pulsatile forces. This mismatch could generate higher salience of error signals that contribute to heightened sensitivity to pulsatile forces.

Limb motion dictates how motor learning arises from arbitrary environmental dynamics

A key idea in motor learning is that internal models of environmental dynamics are internally represented as functions of spatial variables including position, velocity, and acceleration of body motion. We refer to such a representation as motion dependent. The evidence for a motion-dependent representation is, however, primarily based on examination of the adaptation to motion-dependent dynamic environments. To more rigorously test this idea, we examined the adaptive response to perturbations that cannot be well approximated by motion-state: force-impulses--brief, high-amplitude pulses of force. The induced adaptation characterizes the impulse response of the system--a widely used technique for probing system dynamics in engineering systems identification. Here we examined the adaptive responses to two different force-impulse perturbations during human voluntary reaching movements. We found that although neither could be well approximated by motion-state (R(2) < 0.18 in both cases), both perturbations induced single-trial adaptive responses that were (R(2) > 0.87). Moreover, these responses were similar in shape to those induced by low-fidelity motion-based approximations of the force-impulses (r > 0.88). Remarkably, we found that the motion dependence of the adaptive responses to force-impulses persisted, even after prolonged exposure (R(2) > 0.95). During a 300-trial training period, trial-to-trial fluctuations in the position, velocity, and acceleration of motion accurately predicted trial-to-trial fluctuations in the adaptive response, and the adaptation gradually became more specific to the perturbation, but only via reorganization of the structure of the motion-dependent representation. These results indicate that internal models of environmental dynamics represent these dynamics in a motion-dependent manner, regardless of the nature of the dynamics encountered.

Compensation for the intrinsic dynamics of the InMotion2 robot

The InMotion2 and other similarly designed robots, are commonly used for rehabilitation of neurological injuries and motor adaptation studies. These robots are used to simulate haptic environments; however, anisotropy in end-point impedance due to the intrinsic robot dynamics can compromise these experiments. The goal was to decrease the magnitude and anisotropy of the robot impedance using a dynamic compensation algorithm that reduces the forces normally felt by the user during rapid movements. We tested this algorithm with two different methods for real-time calculation of derivatives, a novel quadratic fit method (CQF) and the commonly used backward derivative method (CBD). Six subjects performed a series of point-to-point movements under three conditions (no compensation, CQF, CBD), in different directions at peak speeds of 50, 100 and 150 cm/s. Without compensation, tangential peak-to-peak forces were as large as 69 N in certain directions at the 150 cm/s speed. Both CQF and CBD significantly reduced tangential forces in all directions and speeds. CQF outperformed CBD in the directions with highest intrinsic impedance, reducing tangential forces by 64% in these directions. Compensation also significantly reduced forces normal to the movement direction, with CQF again outperforming CBD in several cases. Anisotropy was assessed by the range of tangential peak-to-peak forces across movement directions. In the no compensation condition, anisotropy was as high as 52.7 N at the 150 cm/s speed, but an average anisotropy reduction of 74% was achieved with CQF. The CQF method can significantly reduce impedance and anisotropy in this class of robot.

Motor memory is encoded as a gain-field combination of intrinsic and extrinsic action representations

Actions can be planned in either an intrinsic (body-based) reference frame or an extrinsic (world-based) frame, and understanding how the internal representations associated with these frames contribute to the learning of motor actions is a key issue in motor control. We studied the internal representation of this learning in human subjects by analyzing generalization patterns across an array of different movement directions and workspaces after training a visuomotor rotation in a single movement direction in one workspace. This provided a dense sampling of the generalization function across intrinsic and extrinsic reference frames, which allowed us to dissociate intrinsic and extrinsic representations and determine the manner in which they contributed to the motor memory for a trained action. A first experiment showed that the generalization pattern reflected a memory that was intermediate between intrinsic and extrinsic representations. A second experiment showed that this intermediate representation could not arise from separate intrinsic and extrinsic learning. Instead, we find that the representation of learning is based on a gain-field combination of local representations in intrinsic and extrinsic coordinates. This gain-field representation generalizes between actions by effectively computing similarity based on the (Mahalanobis) distance across intrinsic and extrinsic coordinates and is in line with neural recordings showing mixed intrinsic-extrinsic representations in motor and parietal cortices.

Adaptation to sequence force perturbation during vertical and horizontal reaching movement-averaging the past or predicting the future?

Several studies conducted during the past decade have suggested that episodic memory is better equipped to handle the future than the past. Here, we consider this premise in the context of motor memory. State-of-the-art computational models for trial-by-trial motor adaptation to constant and stochastic force field perturbations in a horizontal reaching paradigm have shown that motor memory registers a weighted sum of past experiences to predict force perturbation in a subsequent trial. In the current study, we used the standard horizontal reaching movement paradigm and a novel vertical reaching movement paradigm to test motor memory function during adaptation to force fields increasing in magnitude in a simple predictable linear series. We found that adaptation to constant and sequence force fields are similar in vertical and horizontal reaching. For both horizontal and vertical reaching, we found that the expectation in a particular trial was the average of the previous few trials rather than an expectation of a larger perturbation, as would be expected from a simple extrapolation. These findings are not consistent with those of our previous studies on lifting and grasping tasks, in which we found that the grip force is correctly adjusted to the next weight in a series of tasks with gradually increasing weights, thus predicting the future rather than averaging the past. The results of the current study devoted to reaching movements and of our previous study addressing a lifting task suggest that the brain can generate at least two different types of motor representation, either addressing the past in reaching or predicting the future in lifting. We propose that prior experience and the effect of environment's variability are the reasons for the observed differences in expectation during lifting and reaching. Finally, we discuss these two types of memory mechanisms with respect to the distinct neural circuits responsible for lifting and reaching.

Motor learning relies on integrated sensory inputs in ADHD, but over-selectively on proprioception in autism spectrum conditions

The brain builds an association between action and sensory feedback to predict the sensory consequence of self-generated motor commands. This internal model of action is central to our ability to adapt movements and may also play a role in our ability to learn from observing others. Recently, we reported that the spatial generalization patterns that accompany adaptation of reaching movements were distinct in children with autism spectrum disorder (ASD) as compared with typically developing (TD) children. To test whether the generalization patterns are specific to ASD, here, we compared the patterns of adaptation with those in children with attention deficit hyperactivity disorder (ADHD). Consistent with our previous observations, we found that in ASD, the motor memory showed greater than normal generalization in proprioceptive coordinates compared with both TD children and children with ADHD; children with ASD also showed slower rates of adaptation compared with both control groups. Children with ADHD did not show this excessive generalization to the proprioceptive target, but they did show excessive variability in the speed of movements with an increase in the exponential distribution of responses (τ) as compared with both TD children and children with ASD. The results suggest that slower rate of adaptation and anomalous bias towards proprioceptive feedback during motor learning are characteristics of autism, whereas increased variability in execution is a characteristic of ADHD.

Stimulation of the human motor cortex alters generalization patterns of motor learning

It has been hypothesized that the generalization patterns that accompany learning carry the signatures of the neural systems that are engaged in that learning. Reach adaptation in force fields has generalization patterns that suggest primary engagement of a neural system that encodes movements in the intrinsic coordinates of joints and muscles, and lesser engagement of a neural system that encodes movements in the extrinsic coordinates of the task. Among the cortical motor areas, the intrinsic coordinate system is most prominently represented in the primary sensorimotor cortices. Here, we used transcranial direct current stimulation (tDCS) to alter mechanisms of synaptic plasticity and found that when it was applied to the motor cortex, it increased generalization in intrinsic coordinates but not extrinsic coordinates. However, when tDCS was applied to the posterior parietal cortex, it had no effects on learning or generalization in the force field task. The results suggest that during force field adaptation, the component of learning that produces generalization in intrinsic coordinates depends on the plasticity in the sensorimotor cortex.

Skilled throwers use physics to time ball release to the nearest millisecond

Skilled throwers achieve accuracy in overarm throwing by releasing the ball on the handpath with a timing precision as low as 1 ms. It is generally believed that this remarkable ability results from a precisely timed command from the brain that opens the fingers. Alternatively, precise timing of ball release could result from a backforce from the ball that pushes the fingers open. The objective was to test these hypotheses in skilled throwers. Angular positions of the hand and phalanges of the middle finger were recorded with the search-coil technique. In support of the backforce hypothesis, we found that when subjects made a throwing motion without a ball in the hand (i.e., without a backforce), they could not open the fingers rapidly, and they had lost the ability to time finger opening in the 1- to 2-ms range. In addition, relationships were found between the magnitude and timing of hand angular acceleration and finger (joint) extension acceleration. The results indicate that although a central command produced initial hand opening, precise timing of ball release came from a mechanism involving Newtonian mechanics, i.e., hand acceleration produced a backforce from the ball on the fingers that pushed the fingers open. In this mechanism, given the appropriate finger force/stiffness, correction for errors in hand acceleration occurs automatically because hand motion causes finger motion. We propose that skilled throwers achieve ball accuracy by computing finger force/stiffness based on state estimation of hand acceleration and that ball inaccuracy occurs when this computation is imprecise.

Colored context cues can facilitate the ability to learn and to switch between multiple dynamical force fields

We tested the efficacy of color context cues during adaptation to dynamic force fields. Four groups of human subjects performed elbow flexion/extension movements to move a cursor between targets on a monitor while encountering a resistive (Vr) or assistive (Va) viscous force field. They performed two training sets of 256 trials daily, for 10 days. The monitor background color changed (red, green) every four successful trials but provided different degrees of force field context information to each group. For the irrelevant-cue groups, the color changed every four trials, but one group encountered only the Va field and the other only the Vr field. For the reliable-cue group, the force field alternated between Va and Vr each time the monitor changed color (Vr, red; Va, green). For the unreliable-cue group, the force field changed between Va and Vr pseudorandomly at each color change. All subjects made increasingly stereotyped movements over 10 training days. Reliable-cue subjects typically learned the association between color cues and fields and began to make predictive changes in motor output at each color change during the first day. Their performance continued to improve over the remaining days. Unreliable-cue subjects also improved their performance across training days but developed a strategy of probing the nature of the field at each color change by emitting a default motor response and then adjusting their motor output in subsequent trials. These findings show that subjects can extract explicit and implicit information from color context cues during force field adaptation.

Learning from sensory and reward prediction errors during motor adaptation

Voluntary motor commands produce two kinds of consequences. Initially, a sensory consequence is observed in terms of activity in our primary sensory organs (e.g., vision, proprioception). Subsequently, the brain evaluates the sensory feedback and produces a subjective measure of utility or usefulness of the motor commands (e.g., reward). As a result, comparisons between predicted and observed consequences of motor commands produce two forms of prediction error. How do these errors contribute to changes in motor commands? Here, we considered a reach adaptation protocol and found that when high quality sensory feedback was available, adaptation of motor commands was driven almost exclusively by sensory prediction errors. This form of learning had a distinct signature: as motor commands adapted, the subjects altered their predictions regarding sensory consequences of motor commands, and generalized this learning broadly to neighboring motor commands. In contrast, as the quality of the sensory feedback degraded, adaptation of motor commands became more dependent on reward prediction errors. Reward prediction errors produced comparable changes in the motor commands, but produced no change in the predicted sensory consequences of motor commands, and generalized only locally. Because we found that there was a within subject correlation between generalization patterns and sensory remapping, it is plausible that during adaptation an individual's relative reliance on sensory vs. reward prediction errors could be inferred. We suggest that while motor commands change because of sensory and reward prediction errors, only sensory prediction errors produce a change in the neural system that predicts sensory consequences of motor commands.

It is thought that motor adaptation relies on sensory prediction errors to form an estimate of the perturbation. Here, we present evidence that motor adaptation can be driven by both sensory and reward prediction errors. We found that learning from sensory prediction error altered the predicted consequences of motor commands, leaving behind a sensory remapping, whereas learning from reward prediction error produced comparable change in motor commands, but did not produce a sensory remapping. It is possible that the neural basis of learning from sensory and reward prediction errors are distinct because they produce different generalization patterns.

Error correction, sensory prediction, and adaptation in motor control

Motor control is the study of how organisms make accurate goal-directed movements. Here we consider two problems that the motor system must solve in order to achieve such control. The first problem is that sensory feedback is noisy and delayed, which can make movements inaccurate and unstable. The second problem is that the relationship between a motor command and the movement it produces is variable, as the body and the environment can both change. A solution is to build adaptive internal models of the body and the world. The predictions of these internal models, called forward models because they transform motor commands into sensory consequences, can be used to both produce a lifetime of calibrated movements, and to improve the ability of the sensory system to estimate the state of the body and the world around it. Forward models are only useful if they produce unbiased predictions. Evidence shows that forward models remain calibrated through motor adaptation: learning driven by sensory prediction errors.

Linear hypergeneralization of learned dynamics across movement speeds reveals anisotropic, gain-encoding primitives for motor adaptation

The ability to generalize learned motor actions to new contexts is a key feature of the motor system. For example, the ability to ride a bicycle or swing a racket is often first developed at lower speeds and later applied to faster velocities. A number of previous studies have examined the generalization of motor adaptation across movement directions and found that the learned adaptation decays in a pattern consistent with the existence of motor primitives that display narrow Gaussian tuning. However, few studies have examined the generalization of motor adaptation across movement speeds. Following adaptation to linear velocity-dependent dynamics during point-to-point reaching arm movements at one speed, we tested the ability of subjects to transfer this adaptation to short-duration higher-speed movements aimed at the same target. We found near-perfect linear extrapolation of the trained adaptation with respect to both the magnitude and the time course of the velocity profiles associated with the high-speed movements: a 69% increase in movement speed corresponded to a 74% extrapolation of the trained adaptation. The close match between the increase in movement speed and the corresponding increase in adaptation beyond what was trained indicates linear hypergeneralization. Computational modeling shows that this pattern of linear hypergeneralization across movement speeds is not compatible with previous models of adaptation in which motor primitives display isotropic Gaussian tuning of motor output around their preferred velocities. Instead, we show that this generalization pattern indicates that the primitives involved in the adaptation to viscous dynamics display anisotropic tuning in velocity space and encode the gain between motor output and motion state rather than motor output itself.

The nervous system uses nonspecific motor learning in response to random perturbations of varying nature

We constantly make small errors during movement and use them to adapt our future movements. Movement experiments often probe this error-driven learning by perturbing movements and analyzing the after-effects. Past studies have applied perturbations of varying nature such as visual disturbances, position- or velocity-dependent forces and modified inertia properties of the limb. However, little is known about how the specific nature of a perturbation influences subsequent movements. For a single perturbation trial, the nature of a perturbation may be highly uncertain to the nervous system, given that it receives only noisy information. One hypothesis is that the nervous system can use this rough estimate to partially correct for the perturbation on the next trial. Alternatively, the nervous system could ignore uncertain information about the nature of the perturbation and resort to a nonspecific adaptation. To study how the brain estimates and responds to incomplete sensory information, we test these two hypotheses using a trial-by-trial adaptation experiment. On each trial, the nature of the perturbation was chosen from six distinct types, including a visuomotor rotation and different force fields. We observed that corrective forces aiming to oppose the perturbation in the following trial were independent of the nature of the perturbation. Our results suggest that the nervous system uses a nonspecific strategy when it has high uncertainty about the nature of perturbations during trial-by-trial learning.

Reduction in learning rates associated with anterograde interference results from interactions between different timescales in motor adaptation

Prior experiences can influence future actions. These experiences can not only drive adaptive changes in motor output, but they can also modulate the rate at which these adaptive changes occur. Here we studied anterograde interference in motor adaptation – the ability of a previously learned motor task (Task A) to reduce the rate of subsequently learning a different (and usually opposite) motor task (Task B). We examined the formation of the motor system's capacity for anterograde interference in the adaptive control of human reaching-arm movements by determining the amount of interference after varying durations of exposure to Task A (13, 41, 112, 230, and 369 trials). We found that the amount of anterograde interference observed in the learning of Task B increased with the duration of Task A. However, this increase did not continue indefinitely; instead, the interference reached asymptote after 15–40 trials of Task A. Interestingly, we found that a recently proposed multi-rate model of motor adaptation, composed of two distinct but interacting adaptive processes, predicts several key features of the interference patterns we observed. Specifically, this computational model (without any free parameters) predicts the initial growth and leveling off of anterograde interference that we describe, as well as the asymptotic amount of interference that we observe experimentally (R² = 0.91). Understanding the mechanisms underlying anterograde interference in motor adaptation may enable the development of improved training and rehabilitation paradigms that mitigate unwanted interference.

The act of learning one task can not only have direct effects on the performance of other tasks, but it can also affect the ability to learn other tasks. One example of the latter is the phenomenon of anterograde interference in motor adaptation, in which the learning of one adaptation can substantially reduce the rate at which the opposite adaptation can be learned. Here we show that the amount of anterograde interference depends systematically on the strength of a particular component of the initial adaptation rather than on the total amount of adaptation that is achieved. This component of the motor memory evolves more slowly than the overall learning and acts in combination with a quickly evolving component of memory to produce the observed improvement in task performance. We proceed to show that a simple computational model of the interactions between these adaptive processes predicts greater than 90% of the variance in the observed interference patterns, suggesting that this quantitative model may enable the development of improved training and rehabilitation paradigms that mitigate unwanted interference.

Postural constraints on movement variability

Movements are inherently variable. When we move to a particular point in space, a cloud of final limb positions is observed around the target. Previously we noted that patterns of variability at the end of movement to a circular target were not circular, but instead reflected patterns of limb stiffness-in directions where limb stiffness was high, variability in end position was low, and vice versa. Here we examine the determinants of variability at movement end in more detail. To do this, we have subjects move the handle of a robotic device from different starting positions into a circular target. We use position servocontrolled displacements of the robot's handle to measure limb stiffness at the end of movement and we also record patterns of end position variability. To examine the effect of change in posture on movement variability, we use a visual motor transformation in which we change the limb configuration and also the actual movement target, while holding constant the visual display. We find that, regardless of movement direction, patterns of variability at the end of movement vary systematically with limb configuration and are also related to patterns of limb stiffness, which are likewise configuration dependent. The result suggests that postural configuration determines the base level of movement variability, on top of which control mechanisms can act to further alter variability.

Robot-assisted modifications of gait in healthy individuals

This study investigated whether short-term modifications of gait could be induced in healthy adults and whether a combination of kinetic (a compliant force resisting deviation of the foot from the prescribed footpath) and visual guidance was superior to either kinetic guidance or visual guidance alone in producing this modification. Thirty-nine healthy adults, 20-33 years old, were randomly assigned to the three groups receiving six 10-min blocks of treadmill training requiring them to modify their footpath to match a scaled-down path. Changes of the footpath, specific joint events and joint moments were analyzed. Persons receiving combined kinetic and visual guidance showed larger modifications of their gait patterns that were maintained longer, persisting up to 2 h after intervening over-ground activities, than did persons receiving training with primarily kinetic guidance or with visual guidance alone. The results emphasize the short-term plasticity of locomotor circuits and provide a possible basis for persons learning to achieve more functional gait patterns following a stroke or other neurological disorders.

Persistence of motor memories reflects statistics of the learning event

Learning to control a new tool (i.e., a novel environment) produces an internal model, i.e., a motor memory that allows the brain to implicitly predict the behavior of the tool. Data from a wide array of experiments suggest that formation of motor memory is not a single process, but one that is due to multiple adaptive processes with different time constants. Here we asked whether these time constants are invariant or are they influenced by the statistics of the learning event. To measure the time constants, we controlled the statistics of the learning event in a reaching task and then assayed the decay rates of motor output in a set of trials in which errors were effectively removed. We found that prior experience with a rapid change in the environment increased the decay rate of memories acquired later in response to a gradual change in the same environment. Prior experience in an environment that changed gradually reduced the decay rates of memories acquired later in response to a rapid change in that same environment. Indeed we found that by manipulating the prior statistics of the learning experience, we could readily alter the decay rates of a given motor memory. This suggests that time scales of processes that support motor memory are not constant. Rather decay of motor memory is the brain's implicit estimate of how likely it is that the environment will change with time. During motor learning, prior statistics that suggest changes are likely to be permanent result in slowly decaying memories, whereas prior statistics that suggest changes are transient result in rapidly decaying memories.

Consolidation patterns of human motor memory

Can memories be unlearned, or is unlearning a form of acquiring a new memory that competes with the old, effectively masking it? We considered motor memories that were acquired when people learned to use a novel tool. We trained people to reach with tool A and quantified recall in error-clamp trials, i.e., trials in which the memory was reactivated but error-dependent learning was minimized. We measured both the magnitude of the memory and its resistance to change. With passage of time between acquisition and reactivation (up to 24 h), memory of A slowly declined, but with reactivation remained resistant to change. After learning of tool A, brief exposure to tool B brought performance back to baseline, i.e., apparent extinction. Yet, for up to a few minutes after A+B training, output in error-clamp trials increased from baseline to match those who had trained only in A. This spontaneous recovery and convergence demonstrated that B did not produce any unlearning of A. Rather, it masked A with a new memory that was very fragile. We tracked the memory of B as a function of time and found that within minutes it was transformed from a fragile to a more stable state. Therefore, a sudden performance error in a well-learned motor task does not produce unlearning, but rather installs a competing but fragile memory that with passage of time acquires stability. Learning not only engages processes that adapt at multiple timescales, but once practice ends, the fast states are partially transformed into slower states.

Reach adaptation: what determines whether we learn an internal model of the tool or adapt the model of our arm?

We make errors when learning to use a new tool. However, the cause of error may be ambiguous: is it because we misestimated properties of the tool or of our own arm? We considered a well-studied adaptation task in which people made goal-directed reaching movements while holding the handle of a robotic arm. The robot produced viscous forces that perturbed reach trajectories. As reaching improved with practice, did people recalibrate an internal model of their arm, or did they build an internal model of the novel tool (robot), or both? What factors influenced how the brain solved this credit assignment problem? To investigate these questions, we compared transfer of adaptation between three conditions: catch trials in which robot forces were turned off unannounced, robot-null trials in which subjects were told that forces were turned off, and free-space trials in which subjects still held the handle but watched as it was detached from the robot. Transfer to free space was 40% of that observed in unannounced catch trials. We next hypothesized that transfer to free space might increase if the training field changed gradually, rather than abruptly. Indeed, this method increased transfer to free space from 40 to 60%. Therefore although practice with a novel tool resulted in formation of an internal model of the tool, it also appeared to produce a transient change in the internal model of the subject's arm. Gradual changes in the tool's dynamics increased the extent to which the nervous system recalibrated the model of the subject's own arm.

Motor adaptation as a process of reoptimization

Adaptation is sometimes viewed as a process in which the nervous system learns to predict and cancel effects of a novel environment, returning movements to near baseline (unperturbed) conditions. An alternate view is that cancellation is not the goal of adaptation. Rather, the goal is to maximize performance in that environment. If performance criteria are well defined, theory allows one to predict the reoptimized trajectory. For example, if velocity-dependent forces perturb the hand perpendicular to the direction of a reaching movement, the best reach plan is not a straight line but a curved path that appears to overcompensate for the forces. If this environment is stochastic (changing from trial to trial), the reoptimized plan should take into account this uncertainty, removing the overcompensation. If the stochastic environment is zero-mean, peak velocities should increase to allow for more time to approach the target. Finally, if one is reaching through a via-point, the optimum plan in a zero-mean deterministic environment is a smooth movement but in a zero-mean stochastic environment is a segmented movement. We observed all of these tendencies in how people adapt to novel environments. Therefore, motor control in a novel environment is not a process of perturbation cancellation. Rather, the process resembles reoptimization: through practice in the novel environment, we learn internal models that predict sensory consequences of motor commands. Through reward-based optimization, we use the internal model to search for a better movement plan to minimize implicit motor costs and maximize rewards.

Evolution of motor memory during the seconds after observation of motor error

When a movement results in error, the nervous system amends the motor commands that generate the subsequent movement. Here we show that this adaptation depends not just on error, but also on passage of time between the two movements. We observed that subjects learned a reaching task faster, i.e., with fewer trials, when the intertrial time intervals (ITIs) were lengthened. We hypothesized two computational mechanisms that could have accounted for this. First, learning could have been driven by a Bayesian process where the learner assumed that errors are the result of perturbations that have multiple timescales. In theory, longer ITIs can produce faster learning because passage of time might increase uncertainty, which in turn increases sensitivity to error. Second, error in a trial may result in a trace that decays with time. If the learner continued to sample from the trace during the ITI, then adaptation would increase with increased ITIs. The two models made separate predictions: The Bayesian model predicted that when movements are separated by random ITIs, the learner would learn most from a trial that followed a long time interval. In contrast, the trace model predicted that the learner would learn most from a trial that preceded a long time interval. We performed two experiments to test for these predictions and in both experiments found evidence for the trace model. We suggest that motor error produces an error memory trace that decays with a time constant of about 4 s, continuously promoting adaptation until the next movement.