Journal Club Meetings
Niu, et. al., "Neuromorphic Meets Neuromechanics Part I: The Methodology and Implementation" 11/12/2018
Here, the authors present a new approach that can be used to study neuromechanical models. Specifically, they used FPGA's to implement a spiking network model of the alpha motorneurons, muscle, and muscle spidle to control a robotic and cadaveric finger. The authors showed that despite the non-linear inputs (spiking sensory afferents) and non-linear outputs (spiking motor units) the model can accurately perform simple tasks such as posture maintenance under a disturbance load. It seems to show that while the problem is non-linear the redundancy of many noisy copies seems to average out to a smooth signal. The paper is a portion of a 2-part ensemble.
Taylor, J. A., et. al., "Explicit and implicit contributions to learning in a sensorimotor adaptation task" 10/22/2018
This work present an approach to measure explicit learning in the visuomotor rotation task by asking subjects verbally their intended reach direction. Based on the explicitly learning curve measured, here the implicit component of the learning has also been inferred. They conclude that visuomotor adaptation, results from the interplay between explicit learning driven by target error and implicit learning of a forward model driven by prediction error. (A.R.)
Mehrabi, et. al. "Predictive Simulation of Reaching Moving Targets Using Nonlinear Model Predictive Control" 10/29/2018
The authors present a non-linear muscle based controller for predicting planner reaching movements. Interestingly, due to the finite prediction horizon they implement, the model can adapt to changes in target location in between the reach. Their results match the experimental paths but not velocity profiles or EMG data. With a certain cost function they have, their model is an example of a model that produces straight reaches while minimizing muscle effort. (F. Z.)
Pearlmutter, Barak A. "Learning state space trajectories in recurrent neural networks." 10/8/2018
This paper showed how recurrent networks can be thought of as a dynamical system. Furthermore, when training the network the weights learn to represent the appropriate dynamics of the states. For example, a RNN designed to follow a trajectory of a circle, designated certain nodes to define the limit cycle and others to form stable attractors toward the limit cycle. This is an interesting representation of how recurrent networks propagate a signal through time. The same equations derived, here, can be derived through other commonly used methods of solving differential equations such as shooting method, dynamic programming etc. (S.P.)
Gaveau et al., "Direction-dependent arm kinematics reveal optimal integration of gravity cues " 9/17/2018
This paper is about the brain's internal model of gravity. They showed that reaches in a zero-g environment are not directional dependent any more. They introduce Minimum Smooth-Effort model that matches their results and minimizes: jerk + absolute work of muscle torque. They use this result as evidence for the fact that the brain takes advantage of gravity to minimize movement effort. This paper is in contrast with the idea that the brain used internal models for compensation such that a stereotypical trajectory can be maintained. (F.Z.).
Maryam Saleh et al., "Fast and Slow Oscillations in Human Primary Motor Cortex Predict Oncoming Behaviorally Relevant Cues" 9/10/2018
This paper sought to understand the role of two commonly observed frequency bands in motor behavior: the beta band (12-30Hz) and delta band (0.5-2Hz). It has previously been suggested that these bands are associated with held limb postures, but also have been linked to attention. The authors developed an experiment where cues could be either relevant or irrelevant to the instruction of the reach to activate different levels of attention. They found that the beta band amplitude rose and the delta band remained "phase synced" for all cues until they became irrelevant. Thus, they propose that the beta band reflects an attention level, while the delta band reflects an internal metronome to anticipate the next cue. (S.P.)
Werbos, Paul J. , "Backpropagation through time: what it does and how to do it." 7/30/2018
This paper introduces backpropagation through time and explains the differences with basic backpropagation. The paper also gives applications of this method with neural networks and supervised learning as well as neuro-identification and neuro-control (F.Z.).
Fligge et al., "Minimum Jerk for human catching movements in 3D" 7/16/2018
Here, the authors use a catching paradigm to compare human reaches to simulated reaches with a minimum jerk model. Since the boundary constraints are non zero velocity and acceleration, minimum jerk is a curved reach, which closely follows the human trajectories. In addition, the authors found distinct sub-movements in the human catch trajectories indicating what they call an initial "predictive" movement and a later "prospective" or corrective movement for finer manipulation of getting the hand to target. This catching study has interesting implications for point to point movements. It appears that minimum jerk movements are a good approximation for more than just point to point movements, yet it is not clear that the brain uses this control strategy in planning (S.P.).
P. L. Gribble, D. J. Ostry, et al., "Are complex control signals required for human arm movement?" 7/9/2018
They use a two-joint arm model based on the lambda-version of the equilibrium-point hypothesis including 6 muscle forces. They were able to reproduce the results for three other studies with their model but with “simple” commands. They think that models in other studies are too simplified. This study is a example for muscle models. It’s a reference for how muscle models with simple command signals result in realistic movement patterns. (F.Z.)
Zajac FE, "Muscle and tendon: properties, models, scaling, and application to biomechanics and motor control" 4/6/2018
Properties of muscle and tendon (FL, FV curves) were discussed in this review paper alongside a first order model for activation of the muscle as a function of excitation, and a first order model of musculotendon force system as a function of activation, velocity and length. (A.R.)
Dingwell, Mah, Mussa-Ivaldi "Experimentally Confirmed Mathematical Model for Human Control of a Non-Rigid Object." 3/4/2018
They present a mathematical model for control of a spring and a mass spring and mass with a specific BC. They show that the hand chooses paths that are different from min-jerk with bi-phasic velocity profiles. They support the Optimally Smooth Transport (OST) hypothesis which says the hand plans movements to make the object go on the optimally smooth path. (F.Z.)
Stephen Scott "Optimal Feedback Control and the Neural Basis of Volitional Motor Control." 3/5/2018
In general, there are two competing beliefs about how the primary motor cortex controls volitional movements: either M1 controls movements in the domain of high level variables (hand path, global goals) or in terms of low level variables (muscle units, torques). The author recognizes the validity of both ideas, since the M1 has correlations with variables across both domains, and tries to bridge them together with the concept of optimal control feedback. This review clearly describes what about the brain is consistent with optimal control theory (state estimator integration, heavy feedback influence, etc.) and what we need to find to bolster the argument for its presence in the brain.
Marblestone, Wayne and Kording "Toward an Integration of Deep Learning and Neuroscience." 2/5/2018 - 2/19/2018
The recent breakthroughs in deep learning and neural network architecture has bridged a gap between computer science and neuroscience. This paper referenced popular computer algorithms being used to solve complex problems today, theorized ways they could be utilized in the brain and the areas of the brain in which this could occur. The authors proposed three hypotheses: the brain optimizes cost functions, cost functions differ across areas of the brain and across time of development, and biological systems will find efficient solutions to complex problems. This review is a very good organization of ideas on how machine learning has influenced our understanding of learning in the brain and vice versa. (S.P.)
Z. Danziger, F. A. Mussa-Ivaldi "The Influence of Visual Motion on Motor Learning." 1/29/2018
Subjects were able to move a simulated linkage along a geodesic path (straight line) in the geometrical space that they observed their motion. This means that knowledge about the geometric properties of the system can effect the control strategy of the movement.
This study gives evidence for non-straight movements in experimental settings. It is an example of how visual feedback can alter the control strategy of a movement. Most importantly, it is an example of how knowledge about the environment can change the control strategy in a way that the environment is taken into account. (F. Z.)
S Deneve et al. "The brain as an efficient and adaptive learner." 1/19/2018
Due to the “credit assignment” problem (acknowledging certain synapses as reasons for error in the final signal ), backpropagation seems to be biologically improbable. This idea is exaggerated by the fact that our brains have many layers and is highly recurrent. This paper sought an explanation for a biologically relevant method for the brain to assign errors and drive plasticity in the brain. They established that there are two experimentally relevant features that must be maintained in the network: a balance of excitatory and inhibitory inputs and top down error feedback. These features help maintain that the neuron stays close to its threshold voltage and that all neurons receive the same error signal as input to avoid redundancy/promote efficient spiking, respectively. The analysis shows that the networks that contain these features are robust to noise, destructions and perturbations; therefore, it has close similarity with electrophysiological results. (S.P.)
Aaron L. Wong & Adrian M. Haith "Motor planning flexibly optimizes performance under uncertainty about task goals." 12/4/2017
This paper shows that when there is ambiguity in motor goal, subjects perform intermediate movements only at slower speed. Meaning that intermediate reaches are deliberate and beneficial. This is evidence that the motor system executes a single plan and not an average of multiple plans. This paper provides us with a insight into how we prepare actions to achieve desired goals when there is uncertainty about the goals (by F.Z.).
Tenenbaum, Joshua B., and Thomas L. Griffiths. "Generalization, similarity, and Bayesian inference." 11/26/2017
Shepard's universal model of generalization explains this phenomenon from a Bayesian statistics aspect of view. Shepard's ideal model is based on generalization with single encountered stimulus with to a single novel stimulus. Authors in this paper provided an extension to Shepard's idea by introducing the effect of "Size Principle". Using strong sampling as a method of calculating the likelihood, they extended the ideal model for encountering multiple stimuli. This also results in combining the two classically opposing theories of generalization (Shepard's model) and similarity (Tversky's model). (by A.R.)
Sang-Hoon Yeo et al., 2016 "When Optimal Feedback Control is Not Enough: Feedforward Strategies are Required for Optimal Control with Active Sensing" 11/06/2017
They showed that in the presence of a velocity dependent visibility field, subjects make curved reaches that prioritize visibility in the latter portion of the movement. This is a sign that people prefer sensory information towards the end of the movement rather than the beginning. this paper is an interesting behavioral example where people make curved reaches when visual feedback is altered. (by F.Z)
Pouget, "Probabilistic Brains: Knowns and Unknowns" 10/23/2017
Probabilistic inference could be used to explain how multiple uncertain sensory inputs are integrated to make an estimate. The authors explained that whenever basis function is derived from likelihood function the result is a linear probabilistic population code. The author supported the argument with neural data. The author proposes three biological model for this integration: combining multiple sources, marginalization, estimation. These integrations of different sensory information could have great influence on motor behavior and, Bayesian inference could be utilized in modeling it. (by S.H.)
Takakusaki et al., 2004 "Role of basal ganglia-brainstem systems in the control of postural muscle tone and locomotion" 10/06/2017
Basal ganglia diseases have been well studied with their control over voluntary actions, but this is not the case with the automatic control of movement. The authors provided evidence that the basal ganglia serve as an integrator for both voluntary commands from the cortex and automatic commands to the brainstem for pathways used for locomotion and postural muscle tone. They do so by stimulating the substantia nigra (SNr) and studying GABAergic inhibitory connections from the SNr to the PPTN (muscle tone inhibition pathway) and MLR (locomotion initiation pathway). While stimulation of the SNr alone did not change movement behavior, it did have a modulatory effect on movement behavior when costimulated with the PPTN and MLR. Therefeore, the authors provided evidence that the SNr inhibits both the muscle tone inhibition (disinhibition) and locomotion activation. Overstimulating this region created automatic movement deficits similar to those seen in Parkinson's disease patients (ie. decreased gait velocity, stride length, cadence etc), which provides insights as to how to treat the disease. (by S.P)
S. Hocherman and S.P. Wise, 1990 "Effects of hand movement path on motor cortical activity in awake, behaving rhesus monkeys" 09/29/2017
Two Rhesus monkeys were trained to make three types of reaching movements - curved clockwise, straight and counter clockwise trajectories to three different targets. Electrodes were placed and recorded from multiple region from several neurons were read from M1, PMd, PMv and M2 regions during (1) reference period activity, (2) signal-related activity, (3) early set-related activity, (4) late set-related activity (set b), (5) premovement-related activity, (6) late movement-related activity, and (7) return period activity. They found both neurons selective to the trajectory, the target, initial direction and a combination of these parameters (both strict and relaxed criteria). With several neurons showing preference to curved trajectory, this provides evidence that the cortex apart from jus t encoding the movement kinematics, more importantly M1, PMd also encodes these complex aspects of the reach trajectory (by N.V)
Ch 41 - 'Posture', Principles of Neuroscience - Eric R. Kandel; 09/22/2017
Posture is mostly an unconscious effort and much of what we do to keep ourselves upright goes unnoticed. Postural readjustment must be preceded by anticipatory motor actions and can be adapted to suit specific behavioral conditions. Adaptive postural control requires an intact cerebellum and can be learned during locomotion. There are many vestibular and neck reflexes that work together to keep track of the body and head's orientation relative to each other and space. (by S.P.)
Ch 38 - 'Voluntary movements', Principles of Neuroscience - Eric R. Kandel; 08/25/2017
Contralateral Precentral Gyrus or also now known as the Primary Motor Cortex modulates specific neurons, either flex or extend the individual joints of their contralateral limbs and also is the area in which the lowest-intensity electrical stimulation elicits movement. Activity in Individual Neurons of the Primary Motor Cortex Is Related to Muscle Force. We saw how each Cortical Motor Area Receives Unique Cortical and Subcortical Inputs and also learn about the Primary, Supplementary and Pre-motor cortex and how these parts contribute to the different aspects of motor planning. (by N.V.)
Ch 36 - 'Spinal Reflexes', Principles of Neuroscience - Eric R. Kandel
Reflexes are mostly evoked by a peripheral stimulation to the sensory system. The spinal cord is a major site for integration of the reflexes and the central commands. Unlike conventional definitions reflexes are not stereotyped and they are modifiable. Examples of well-known reflexes are discussed in this chapter such as flexion-withdrawal, stretch, and proprioceptive reflex. Muscle spindles act like length sensors in muscles and can be used as a sensory input to sense the relative position of body segments. Golgi tendon organs have the function of tension sensors in muscles and are always kept in an optimal length of functionality by motor neurons. Proprioceptive reflex signals contribute to adjusting motor output according to the state of the limbs. Supraspinal pathways can be involved in reflexes of limbs. (by A.R.)
Ch 34 - 'The Motor Unit and Muscle Action', Principles of Neuroscience - Eric R. Kandel
The main goal of the body, and what makes animals distinguishable from plants, is to make goal directed movements. The Contractile Machinery of Muscle Fibers is Organized into Sarcomeres and Cross Bridges. Sarcomeres contain thick and thin filaments, which interact in a special way to produce muscle force (uses ATP to reload). There are three types of motor units, slow twitch, fast twitch, and fatigue resistance fast twitch (these depend on type/duration of usage). Motor units are recruited in a specific order to allow for dexterous movements. Also, motor unit firing rate increase with load/demand. Finally, co-contraction of opposing muscles can resist disturbances and/or provide stability.
Ch 33 - 'Organisation of movements', Principles of Neuroscience - Eric R. Kandel
Key concepts - Motor cortex and the different senses it uses for motion planning. Sensory delays. Reflex, repetitive motor sequences, goal directed voluntary movements. Feedback vs feed forward control. Ball catching expn. 3 Psycho physical principles governing motor action: (1) Brain represents movement as an abstract motor action independent of the end effector (e.g. Motor Equivalence, movement scaling) (2) Time for feedback from rich sensory information takes more time (3) Accuracy of movement is inversely correlated with speed of motion. Motor System’s hierarchy - Peripheral → Brain stem → Cortex. Functions of Basal Ganglia and Cerebellum. Spinal Motor Neurons. Function, cross sectional arrangement of Spinal tracts and motor pathways #movement #motorpathways #neuralcircuitry (by N.V)
"Rhythms of the Hippocampus" 2016 Laura Lee Colgin
The author describes in detail several types of waveforms encountered int he hippocamous. Specifically, she describes how these waveforms influence a special pyramidal cell in the hippocampus called place cells. These cells respond with higher firing rate when near a specific area in a room. Similar, to visual fields, a place field describes the boundary with which these place cells fire. The three waveforms discussed were theta waves, short-wave ripples (SWR) and gamma waves. The theta wave is large amplitude, slow wave resulting from a large coordination of firing. Certain place cells fire heavily at different points on the theta wave, and this special sequence of place cell firing is believed to be how we remember where we've been or where to go. Theta waves are also believed to coordinate certain places with other senses (light,smell,touch) to store a memory. SWRs are seen shorter waves seen directly after theta waves, and they occur at a much higher frequency (>10x). Shortly after a path through certain place fields have been taken, the exact sequence of place cells will fire repeatedly but on a shorter time scale. It is believed that this "replay" is how memory is stored to the cortex/remembered at later times. Finally, gamma wave function has been relatively inconclusive, but they may play an intermediate functions between sensory acquisition and storage of the memory. (by S.P.)
Henk J Groenewegen "The Basal Ganglia and Motor control" 2003
The author here describes in detail the functional anatomy of Basal Ganglia - the parts striatum, pallidum, subthalamic nucleus, substantia nigra its influence in the motor pathways along with the neuro transmitters. It is seen that the basal ganglia has two pathways involved in the selection, sequencing and execution of motor commands - the direct and indirect pathway. A higher activity in the ‘direct’ striatal output pathway (prefrontal thalamocortical systems) is considered to be associated with increased motor or behavioral output and facilitates movements and the absence of which would result in discontinuous or intermittent movements. Whereas the activity in indirect striatal pathway (along subthalamic projections) suppresses unintentional movements. The role of Dopamine added to its effect on the motivational pathway is hypothesized to help in the correction of movements with ‘gating’ the activation on spiky neurons. #BasalGanglia #movement_correction (by N.V)
Jennifer L. Raymond, Stephen G. Lisberger, Michael D. Mauk “The Cerebellum: A Neuronal Learning Machine?” (1996)
For two specific motor behaviors, they showed how the cerebellar cortex and the cerebellar nuclei play a role in plasticity, timing and transferring learning. The neural mechanism used in the cerebellum for these two behaviors (classic conditioning eyelid response, vestibulo-ocular reflex), may represent all forms of cerebellum-dependent learning, such as arm movements. (by F.Z) #cerebellum #Neural_Learning #eyelid_response #VOR
David A. Robinson “The Oculomotor Control System: A Review” (1968)
This is a review paper on Oculomotor system studies. It explains the technology involved for measuring eye movement at that time. The paper talks about the functions and the anatomical structures for saccadic, smooth pursuit, vergence and the vestibular systems. The overall transfer function between motor neurons and eye movements are also discussed in this paper. (by F.Z) #oculomotor #control_theory
Alon Fishbach, Stephane A. Roy Christina Bastianen, Lee E. Miller James C. Houk “Deciding when and how to correct a movement: discrete sub movements as a decision making process” Exp Brain Res (2007)
They studied the asymmetric multi-peaked velocity profiles observed from the one-dimensional step-tracking task on a screen by supination/pronation of the forearm made by trained monkeys. The profile was decomposed into discrete sub-movements by “soft symmetry method” namely the Primitive movements, Overlapping sub movements and Delayed sub movements. From different analyses of the velocity profiles it is evident that the corrective sub movements were initiated when the probability distribution of the predicted end-point from the continuously accumulated information is statistically different from the target’s location. They also showed that the latency in onset of the following sub movements are proportional to the amplitude of the previous primitive and the ‘extent to go’ of the reach. Thus, these evidences claim that there is not only a continuous feedforward controller but it triggers a discrete execution when an error threshold is met. (N.V) #velocityprofile #discretecontroller
I. S. Howard, D. W. Franklin and D. M. Wolpert "Gone in 0.6 Seconds: The Encoding of Motor Memories Depends on Recent Sensorimotor States"
Subjects can learn interfering force fields only if dynamic (active or passive) cues are available prior to the movement. The cue should be present in a time window of 600ms prior to the movement. This means that subjects are able to learn a force field if there is a special movement associated with that force field. This paper is evidence for the fact that dynamic motor learning is not a map through current states to motor output but recent history of the previous states have fundamental role in learning. (by F.Z.) #interference #forcefieldadaptation #motormemory
Mattar, Andrew AG, and David J. Ostry. "Generalization of dynamics learning across changes in movement amplitude."
The authors showed that in dynamics learning, decreasing the amplitude of a test reach relative to training allows generalization, while increasing the amplitude does not. The maximum threshold at which generalization is possible corresponds to the moment at which velocity increases beyond that experienced in test trials. Accordingly, generalization is based on experienced dynamics, rather than the brain learning exactly how a force field works and building a model. This corroborates the findings of the Mattar & Ostry 2007 paper. Since the experience gained by reaching for one target applies minimally to reaching for another target 90° away, it makes sense that there is little to no generalization. Thus, this paper builds the theory that generalization is based on past experience (like a lookup table), rather than constructing a model. Additionally, interleaving test targets with training targets artificially impacts generalization by making subjects no longer naïve, and should be avoided. (by T.A.)
Mattar, Andrew AG, and David J. Ostry. "Modifiability of generalization in dynamics learning."
Generalization of dynamics learning in hand movement is highly localized to the very nearest movements and does not benefit from an extensive distribution of movement experience throughout the working space, however, it depends on interpolating to the local instances. Therefore, there is just a narrow dynamics learning generalization.The results of this study suggest that the process behind dynamics learning relies on performing an interpolation rather than building a model. (by A.R)
Todorov, Emanuel. "Direct cortical control of muscle activation in voluntary arm movements: a model"
The author displayed that a model of motor cortex control can qualitatively represent many recent neural data findings and reaching movement behaviors. If this model is true, the cortex could be a low-level controller directly determining the muscle activations needed to make a movement. This could be a novel representation of the general interaction between motor cortex activity and the kinematics of the hand, which coexists with the findings of Georgopoulos and other studies. (by S.P.)
Georgopoulos et al. "Neuronal population coding of movement direction"
The experiment studied the motor cortex activity of Rhesus monkeys, using microelectrode penetrations, when they performed reaching movements in 3D space. The movement direction of the monkey’s arm was correlated with the rate of firing of the neurons and their vectorial sum known as the population vector. As this population vector predicted accurately the movement vector, it can be understood that the motor cortex activity represents the kinematics of the arm’s reaching movement. Thus, the research gave more insights on how the motor cortex represents a step, of the entire control process, before the brain commands the muscles to perform the movement. (by N.V)
Max Berniker and Konrad Kording "Estimating the Relevance of World Disturbances to Explain Savings, Interference and Long-Term Motor Adaptation Effects "
This study presents a new model that demonstrates properties of motor adaptation such as short-term/long-term savings, inference and spontaneous rebound. They suggest that the CNS both estimates the sources and the relevance of disturbances. This model could be an answer to the deficiencies of the linear model. (by F.Z.)
Max Berniker and Konrad Kording "Estimating the sources of motor errors for adaptation and generalization"
The authors showed that a model that estimates the source of motor errors (either world or body) can explain adaptaion and generalization in reaching movement experiments. Instead of explaining adapataion and generalization in terms of intrinsic or extrinsic coordinate frames, they provide evidence for an internal probabilistic model that updates limb and world paramaters. The CNS may be constatnly updating these parameters to understand adapt to disturbances and generalize to other workspaces or limbs. (by S.P.)
Wang, Jinsung and Robert L. Sainburg "Interlimb Transfer of Novel Inertial Dynamics Is Asymmetrical "
The experiment studies two groups of dominantly right handed people, making reaching movements to intrinsic targets with an external mass attached to their forearm. They show that the adaptation with the dominant arm allowed for more accurate movement in their non-dominant arm than the group which had their adaptations in the non-dominant arm and tried moving with their dominant hand. This bolsters the case for the non-dominant arm controller, having better access to the memory resources where the movement information obtained during dominant arm adaptation is stored. Adaptations to visuomotor and dynamic transformations having different generalizations also point out that these adaptations may involve distinct neural processes. (by N.V.)
Emilio Bizzi et al. "Posture Control and Trajectory Formation During Arm Movement"
During single joint, long reaching movements of the forearm for both intact and deafferented monkeys, the acceleration of the limb does not rapidly approach a constant value, but gradually approaches a constant value well after (~400 msec) the beginning of EMG activity. In addition, intact and deafferented monkeys both returned to an intermediate position between the start and end points of the reach, after being pushed towards or moved to the final position. This suggests that the CNS has programmed a slow shift in equillibrium point along the trajectory. This is evidence against the "final position control" hypothesis and in favor of a "series of equillibrium points" throughout a movement.
Paul Dizio and James R. Lackner "Motor adaptation to Coriolis Force Perturbations of Reaching Movements: Endpoint but not Trajectory Transfers to the Nonexposed Arm"
Making reaches in a null field, after adapting to a coriolis force field with the right hand, subjects make straight but deviated reaches with their left hand. Afterwards, when they make a reach with their right hand, they make curved reaches but with large endpoint errors. This means that endpoint control and path control are independent from each other. These results are in contrast with the equilibrium theory (by F.Z.)
James R. Lackner and Paul Dizio "Rapid Adaptation to Coriolis Force Perturbations of Arm Trajectory"
The author demonstrated that people can learn to adapt to relatively new types of forces, without any aid from visual or haptic feedback. They observed the reaching movements of subjects in a rotating room, under the influence of Coriolis force. Experiments were devised to have both touch and no touch movements showing different end point and trajectory errors, suggesting that the brain has more than one controller for reaching movements (by N.V.)
Ali Farshchian, et al. "Sensory Agreement Guides Kinetic Energy Optimimzation of Arm Movements during Object Manipulation"
People who saw and felt a double pendulum made curved reaches, while those who saw a cursor or felt a point mass made straight reaches. This means that subjects choose the efficient path when the sensing modalities are in agreement but, go straight otherwise. This shows that visual feedback is a critical piece of information for the motor system and the standard visual feedback provides strong bias towards straight movements (by F.Z.). #forcefieldadaptation #visualfeedback #efficient
Franklin, David W., et al. "Visual feedback is not necessary for the learning of novel dynamics."
The authors showed that subjects can learn to make accurate and straight reaching movements in alternative fields with and without vision. For point to point reaching movements, subjects will learn to make straight reaches; however, this may be conditioned upon to fact that a subject can use visual feedback to return the hand to the starting position. If true, then proprioception, not vision, is the primary feedback used for learning.
Max Berniker, et al. "Simplified and Effective Motor Control Based on Muscle Synergies to Exploit Musculoskeletal Dynamics"
The authors developed a low dimensional control model, based on muscle synergies and compared its performance to other conventional control models. A low dimensional controller can be successful in simplifying control while maintaining accuracy, it is a behavior specific controller that does not generalize well to all behaviors. This would mean that the CNS uses task relevant low-dimensional controller and muscle synergies to simplify high dimensional control problems of movement and gives an explanation for perhaps why synergies would exist.
Valero-Cuevas et al. "Challenges and New Approaches to Proving the Existence of Muscle Synergies of Neural Origin"
They demonstrated feedback and feed forward related muscle synergies using cadaveric human hand and computational model of the human leg, simulating low dimensional muscle activations (EMG responses) creating the appearance of muscle synergies even in the absence of a specific neural controller. Hence this supports the claim that the low dimensionality of the muscle activation and constraints are not only of neural origin but also a combinational effect of bio-mechanical constraints. (by N.V.)
Mathew C. Tresch, et al. "The Construction of Movement by the Spinal Cord"
By looking at a number of motor behaviors and muscle EMG records they show that the motor behavior can be explained by a low dimensional method. This means that the muscle synergy method can explain motor behavior in vertebrate spinal cord. Their results could explain the simplification method that the spinal cord uses to control movements. (by F.Z.)
Mathew C. Tresch, et al. "The Case For and Against Muscle Synergies"
The author presented evidence for and against the muscle synergy hypothesis. Although, many studies have provided evidence that the CNS can create movements with a low dimension of synergies, these experiments only analyze a limited range of movements or behaviors. A study that encompasses all possible movements of an organism and still manages to find a low dimension of muscle synergies would provide very strong evidence for this hypothesis. If the muscle synergy hypothesis were true, then it would show that the brain actively reduces the degrees of freedom to reduce computational load during movement. (by S.P.)
Robert L Sinburg et al. “Control of limb dynamics in Normal Subjects and Patients without Proprioception”
The experiments studied both normal subjects and patients with large-fiber sensory neuropathy (lack the sense of proprioception) moving their arms with slicing motions in different directions. It was observed that the patients had greater directional errors and conspicuous drifts at the end of movements. The authors infer that this is mainly due to the failure to compensate for their inertial variation along the motion of their arms and poor inter-joint coordination as result of impaired feedback. (by N.V.)
Gordon, James, et al. "Impairments of Reaching Movements in Patients Without Proprioception. I. Spatial Errors"
The authors studied multi-directional reaching movements between healthy subjects and patients lacking proprioception, by analyzing the latter's failure to reach targets. They provided evidence that directional dependent errors by patients are caused by a failure to account for the biomechanics of the limb, however the evidence was not enough to rule out other factors such as noisy signals or the lack of mechanical ability to know when they have stopped. A thorough understanding of why patients fail to make simple reaching movements could help provide better rehabilitation for patients and help us understand which contributions to movement are feedforward control vs. feedback control. (by S.P.)
P. Ramkumar et al., "Chunking as the result of an efficiency computation trade-off"
Sequence reaching in monkeys is studied. The authors show that the optimal motor control problem can be reframed as a series of chunks and local optimization within those chunks. This could be a possible strategy for the motor system in learning reaching movements. (by F.Z.)
M. Conditt, F. Gandolfo & F. Mussa-Ivaldi, "The motor system does not learn the dynamics of the arm by rote memorization"
The authors provide evidence that after adapting to a force field making point-to-point reaches, circle movements in the null and force field are similar to those when circle movements are practiced directly. Thus the motor system does not merely memorize how to make movements, but rather learns general information about them. In turn this is strong evidence for an internal model that infers the dynamics of our reaches. (by M.B.)
L. N. Gonzalez Castro, M. A. Smith "The Binding of Learning to Action in Motor Adaptation"
They are interested in adaptation rates and the way they are improved. They study Plan Reference Learning versus Motion Reference Learning model and compare these two theories with actual data. They found that the experimental results are closer to MRL. Their results show the importance of feedback errors on the model formed in the CNS. Which indicates the crucial effect of feedback during adaptation (by F.Z.)
T. Flash, "The Control of Hand Equilibrium Trajectories in Multi-Joint Arm Movements"
Human arm stiffness and equilibrium-point trajectory during multi-joint movement paper showed that stiffness of the limb is smaller than previously presented and that it changes dramatically throughout the movement. It is important to know because we can better understand the internal model of the brain based on presented data (by U.S.)
R. C. Miall, P. N. Haggard, "The curvature of human arm movements in the absence of visual experience"
They found that blind subjects make straighter reaches than the ones wearing blind folds. However, their perception of curvature is not statistically distinguishable. So this paper again (like the paper by Sergio) shows that vision is not a sufficient explanation for making straight reaches. (by F.Z.)
M. Kawato, "Feedback-Error-Learning Neural Network for Supervised Motor Learning"
The author presented a new learning rule. This might be helpful when trying to understand unsupervised learning i.e. when we don't know what the desired output should be. (by V.S.)
K. Kording, D. Wolpert, "Bayesian integration in sensorimotor learning"
When making simple decisions subjects make correct use of both the likelihood and the prior distribution of events, consistent with Bayesian integration. It could be the case the the brain makes use of similar Bayesian computations in many other processes. (by M.B.)
M.Berniker, M. Voss, K. Kording, "Learning Priors for Bayesian Computations in the Nervous System"
The paper shows how Bayesian statistics can explain human behavior by combining prior and likelihood. Is it important in optimizing rational decision making and minimizing uncertainties. (by U.S.)
Lauren E. Sergio, Stephan H. Scott, "Hand and Joint Paths During Reaching Movements With and Without Vision"
The aim of this paper is to see the effect of vision on limb trajectories.
The results show that subjects that were congenitally blind make straighter movements compared to blindfolded subjects.
They make a conclusion that there is a powerful tendency of making straight hand trajectories, even when no vision is involved. They also mention that visual feedback itself is not sufficient to explain why subjects tend to generate straight hand paths during movements. (by F.Z.)
Shadmehr & Mussa-Ivaldi, "Adaptive Representation of Dynamics during Learning of a Motor Task"
The authors addressed whether we adapt to change in dynamics, the presence of a kinematic plan, generalizing of the internal model, what the coordinates of the internal model are, and whether they could simulate the results of the experiment. They found that we do adapt, there exists a kinematic plan, the internal model is intrinsic and can be generalized over different workspaces. Also, the results could be simulated. The authors studied the aftereffects, tested how you adapt, and adaptation in different workspaces. They also showed evidence for an intrinsic coordinate based system. (by A.C.)
Maurice A. Smith et. al., "Interacting Adaptive Processes with Different Timescales Underlie Short-Term Motor Learning"
In this study it is shown that a two state, gain-independent, multi-rate model is a better representation for motor adaptation simulations. The spontaneous rebound in channel trials could be explained by this model.
This paper is the first study on spontaneous rebound in motor adaptation. Their model shows that in motor learning many time scales are involved. (by F.Z.)
Fritzie Arce et. al., "Differences in Context and Feedback Result in Different Trajectories and Adaptation Strategies in Reaching"
In this paper the authors found that prior history of learning influences the control policy during adaptation and continuous visual feedback is needed to make straight reaches in a force field.
Their experiment is important to know that under some certain conditions subjects do make curved trajectories in planer reaching movements. (by F.Z.)
John W. Krakauer, Maria-Felice Ghilardi and Claude Ghez, "Independent learning of internal models for kinetic and dynamic control of reaching"
The main finding that was supported by the results of the experiments is that learning of another model with conflicting sensorimotor mappings interferes with the memory of previously learned models. This concept improves current computational methods as it suggests including both kinematic and dynamic sources of error, instead of combining both performance errors. (by M.D.)
Sarah E. Criscimagna-Hemminger, Opher Donchin, Micheal S. Gazzaniga and Reza Shadmehr, "Learned Dynamics of Reaching Movements Generalize From Dominant to Nondominant Arm"
The purpose of the paper is to investigate inter-limb generalization. The goal is to discover the type of coordinate this generalization is based on and the neural basis of it. The result shows that generalization occurred only from the dominant to the non-dominant hand and it is based on an extrinsic coordinate. Moreover, inter-hemisphere communication does not depend on the corpus callosum. The result is useful to understand the ability of brain in executing what it has learned. (by F.Z.)